Multi-TINE weeder

Abstract

A method for eradicating weeds may include positioning a platform amongst a crop; capturing images of at least a portion of the crop proximal to the platform; processing the images to determine at least locations of wanted and unwanted plants in the images and to transform the locations, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; individually moving electric treatment modules that are supported by the platform in at least a z- direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to remain distal from a wanted plant and contact or be proximal to an unwanted plant; and individually actuating each of the electric treatment modules while at respective actuation locations thereby to deliver an eradication treatment to the unwanted plants. Related systems are disclosed.

Description

MULTI-TINE WEEDERCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to United States Provisional Patent Application Serial No.63 / 748,929 filed on January 23, 2025 entitled “TINE WEEDER”, and to International (PCT) Patent Application No. PCT / CA2025 / 050641 filed on April 30, 2025 entitled “SYSTEMS AND METHODS FOR THE DETECTION AND ERADICATION OF WEEDS”, the contents of each of which are incorporated herein by reference.FIELD OF THE DISCLOSURE

[0002] This disclosure relates generally to computational and electrical systems and methods, and more particularly to computational and electrical systems and methods for the detection and eradication of unwanted plants such as weeds.BACKGROUND OF THE DISCLOSURE

[0003] In the next three decades, the global population is projected to rise to over 10 billion. This substantial growth will place a significant demand on global food production, compelling farmers to rapidly and significantly increase their output. Weeds represent one of the most major threats to agricultural yield. Weeds host a range of problematic insects and crop diseases, drain essential resources such as water and nutrients from neighboring crops and in some cases, can release chemicals into the soil to inhibit the growth of nearby plants. Larger weeds can overshadow and block sunlight to crops.

[0004] Plant classification poses a significant challenge due to the spectral similarities shared among different plant species. Existing mapping methods in fields typically presume that host plants (e.g. crops) are sown in rows. These methodologies employ line detection techniques to categorize plants in a row as host plants and those outside the seeding lines as weeds.

[0005] Current agricultural practices predominantly rely on herbicides and manual weeding as the main strategies for weed control. Recent years have witnessed a surge in herbicide-resistant weeds, forcing farmers to increase their use of herbicides or resort to stronger versions. Historically, farmers have applied agrochemicals, including herbicides, fungicides, pesticides, and fertilizers, across their fields in large, indiscriminate quantities.

[0006] Herbicides have been shown to cause adverse effects on biodiversity, soil health, air quality, and water systems. Their impact extends beyond the fields, affecting the habitats of insects, birds, and aquatic species. Herbicides can disrupt soil microbiota, which are essential agents in nutrient cycling, soil structure formation, and plant growth. Water contamination is another serious concern, as herbicides can seep into water bodies via agricultural runoff or leach into groundwater, posing threatsto aquatic life and compromising the quality of drinking water. Additionally, herbicides can contaminate the air, particularly during application when spray drift may occur.

[0007] Despite the historical reliance on herbicides for weed control, an increasing number of weeds are developing resistance. This trend, combined with the known toxicity and environmental issues related to herbicide use, as well as the high costs and wastage associated with indiscriminate field spraying, underscores the impetus for alternative weed control methods. This issue is particularly pertinent for vegetable farmers, for whom weeding constitutes a significant expense.

[0008] Non-chemical weeding methods, such as manual weeding or hand weeding, have been explored as alternatives. However, they present their own set of challenges. Mechanical weeding, as it has been done in the past, can involve substantial soil disturbance and typically utilizes various types of machinery to physically remove or damage weeds. This method can disrupt the soil structure, potentially leading to soil erosion and the loss of important nutrients. It may also affect the beneficial microorganisms that inhabit the soil, which are vital for nutrient cycling and supporting plant growth. Additionally, weeds have a high chance of re-rooting after mechanical treatment. Laser weeding, another alternative, tends to be limited to exterminating early-stage weeds and is ineffective against late-stage weeds.SUMMARY OF THE DISCLOSURE

[0009] In an example, there is provided a system for electrocuting unwanted plants amongst a crop. The system may include a platform positionable amongst the crop; an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform; a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; an electric treatment subsystem supported by the platform, the electric treatment subsystem comprising an array of electric treatment modules extending across the platform, each of the electric treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually electrically actuated to deliver a time-limited electrocution treatment; and a control subsystem operable to move each of the electric treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually electrically actuate each of the electric treatment modules while at the respective actuation location thereby to deliver the time-limited electrocution treatment to the unwanted plants.

[0010] In another example, there is provided a method for eradicating weeds amongst a crop. The method may include positioning a platform amongst the crop; capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform; processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; individually moving electric treatment modules that are supported by the platform in at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; and individually actuating each of the electric treatment modules while at respective actuation locations thereby to deliver an eradication treatment to the unwanted plants.

[0011] In another example, there is provided a system for mechanically harming unwanted plants amongst a crop, the system comprising: a platform positionable amongst the crop; an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform; a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; a mechanical treatment subsystem supported by the platform, the mechanical treatment subsystem comprising an array of mechanical treatment modules extending across the platform, each of the mechanical treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually actuated to execute a mechanical harming action; and a control subsystem operable to move each of the mechanical treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually actuate each of the mechanical treatment modules while at the respective actuation location thereby to execute the mechanical harming action to the unwanted plants.

[0012] In another example, there is provided a method for eradicating weeds amongst a crop, the method comprising: positioning a platform amongst the crop; capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform; processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; individually moving mechanical treatment modules that are supported by the platform inat least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; and individually actuating each of the mechanical treatment modules while at respective actuation locations thereby to deliver a mechanical harming action to the unwanted plants.

[0013] Other aspects are described in detail herein.BRIEF DESCRIPTION OF THE FIGURES

[0014] Examples will now be described with reference to the appended figures, in which:

[0015] FIG. 1A is an image, such as a frame of digital video, of a portion of a field in which a crop plant has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “CROP 22 (SEMI -AUTO)” while the pixels corresponding to the crop plant itself have been changed in color as a contour mask.

[0016] FIG. IB is an image, such as a frame of digital video, of a portion of a field in which a weed plant has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “WEED 8 (SEMI -AUTO)” while the pixels corresponding to the weed plant itself have been changed in color as a contour mask.

[0017] FIG. 1C is an image, such as a frame of digital video, of a portion of a field in which a weed plant has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “WEED 273 (SEMI-AUTO)” while the pixels corresponding to the weed plant itself have been changed in color as a contour mask.

[0018] FIG. ID is an image, such as a frame of digital video, of a portion of a field in which a crop plant has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “CROP 428 (SEMI-AUTO)” while the pixels corresponding to the crop plant itself have been changed in color as a contour mask.

[0019] FIG. 2 is a schematic diagram showing an applicator unit - an electric treatment module or “tine” - supported by a platform and an earth unit also supported by the platform, forming an electric circuit through a weed.

[0020] FIG. 3A is an image, such as a frame of digital video, of a portion of a field in which a weed plant has been identified, and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “WEED 516 (SEMI -AUTO)” and showing contact points on the weed for contact by an electrode.

[0021] FIG. 3B is an image, such as a frame of digital video, of a portion of a field in which a weed plant has been identified, and its location in the image delineated with a rectangular graphical overlay as a bounding box that is labelled / annotated “WEED 508 (SEMI- AUTO)” and showing contact points on the weed for contact by an electrode.

[0022] FIG. 4 is a graphical representation of a portion of a field in which two overlapping portions of a weed plant are depicted and showing contact points on the weed for contact by an electrode.

[0023] FIGS. 5 and 6 are graphical representations of accessed and grid images of a portion of a crop, with weed plants and a crop plant identified.

[0024] FIG. 7 is a 2D graphical representation of an eradication map showing electrocution points and an avoidance area.

[0025] FIG. 8 is a 3D (three-dimensional) graphical representation of an eradication map showing electrocution points and an avoidance volume representing a 3D subregion in the overall 3D region that movement of an electric treatment module, as it traverses 3D space by movement of the platform to which it is fixed in the x-y direction and further controlled to move in a z-direction, according to the eradication map, should avoid.

[0026] FIG. 9 is a 3D graphical representation of an eradication map showing electrocution points, an avoidance volume, and a path on which the electrocution points sit and that goes around (i.e. avoids) the avoidance volume.

[0027] FIG. 10 is a top perspective view of components of a system for electrocuting unwanted plants amongst a crop, with a cover of the system removed to show components beneath, according to an example.

[0028] FIG. 11 is a rear view of components of the system for electrocuting unwanted plants amongst a crop as in FIG. 10, according to an example.

[0029] FIG. 12 is a side view schematic of the system of FIGS. 10 and 11 for electrocuting unwanted plants amongst a crop, according to an example.

[0030] FIG. 13 is a flow diagram showing various stages in a process of detecting and eradicating weeds using the system of FIG. 12, according to one example.

[0031] FIG. 14 is a bottom perspective view of an actuator and an electric treatment module, or “tine,” supported by the actuator;

[0032] FIG. 15 is a side elevation view of the actuator and tine of FIG. 14;

[0033] FIG. 16 is a top view of a transmission system for the actuator and tine of FIG. 14;

[0034] FIG. 17 is a flowchart depicting a method for eradicating weeds amongst a crop, according to an example;

[0035] FIG. 18 is a schematic diagram showing a hardware architecture of a computing system; and

[0036] FIG. 19 is a diagram showing a representation of a crop with wanted and unwanted plants that an arrangement of treatment modules, such as tines, may address for treatment.DETAILED DESCRIPTION

[0037] The proposed design intends to eliminate the need for herbicides by providing an alternative to eliminate weeds with high power electrical energy.

[0038] Patent Cooperation Treaty (PCT) Application No. PCT / CA2025 / 050641 to Tao et al. (“Tao 1”) discloses various examples of systems for electrocuting unwanted plants amongst a crop. The systems include a platform positionable amongst the crop, an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform, a processing subsystem processing the images received from the imaging subsystem to determine at least locations of unwanted plants in the images and to transform the locations in the images into actuation locations in a three-dimensional region proximal to the platform, at least one electric treatment module supported by the platform, each of the at least one electric treatment module controllable to be moved with respect to the platform to actuation locations within the three-dimensional region and to be electrically actuated to deliver a time-limited electrocution treatment, and a control subsystem operable to move the at least one electric treatment module to one or more of the actuation locations in the three-dimensional region thereby to contact or be proximal to unwanted plants, the control subsystem operable to electrically actuate the at least one electric treatment module while at the actuation locations thereby to deliver the time-limited electrocution treatment to the unwanted plants.

[0039] Certain examples provided in Tao 1 may have just one electric treatment module that is mechanically moved through three dimensions with respect to the platform - that is, in x, y, and z directions with respect to the platform - to actuation locations throughout the whole of the three-dimensional region. Examples of systems in accordance with Tao 1, with two or more electric treatment modules, but still a small number of modules, may involve mechanically moving each electric treatment module in all three directions with respect to the platform to respective actuation locations within subregions of the whole three-dimensional region. While an electric treatment module being movable in all three directions with respect to the platform to reach a given actuation location, or to avoid any avoidance location, is useful, the mechanical subsystem for moving in the three directions can be complex and can be accordingly subject to significant wear and tear. It may be useful to provide alternative examples of systems for electrocuting unwanted plants amongst a crop that do not have as much complexity and / or do not incur as much wear and tear.

[0040] The present application presents a system for electrocuting unwanted plants amongst a crop that employs an array of electric treatment modules extending across the platform. In some examples, each of the electric treatment modules have a fixed x-y position with respect to the platform, at least during a given treatment session. With such examples, instead of being moveable in the x ory directions with respect to the platform, the electric treatment modules are individually controllable to be moved in only a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region. In this manner, the mechanical movement of each electric treatment module in only the z-direction can be done with reduced complexity and without incurring as much wear and tear than the Tao 1 configuration might do, or to simply be applied in different environments. These electric treatment modules, or “tines”, are preferably as densely packed as ispractical in the array so as to enable as high-resolution electrocution and avoidance as possible while also providing thorough coverage of the crop.

[0041] It will be noted that, with Tao 1, a system having one electric treatment module movable extensively in three dimensions within the whole of the three-dimensional region with respect to the platform, or an electric treatment module movable extensively in three dimensions with a respective subregion of the whole three-dimensional region can be moved in any or all of the x, y, and z directions to avoid physical contact with wanted plants or other objects that it is not desirable to contact. However, in examples described herein wherein an electric treatment module is fixed in an x-y position with respect to the platform, it will avoid contact with wanted plants only by lifting sufficiently high in the z-direction to an avoidance location; otherwise, the electric treatment module may advance into physical contact with a wanted plant as the platform advances through the crop, possibly physically damaging the wanted plant through physical contact, even though electrocution may not be applied at the time. An array of such electric treatment modules may present collectively to the crop as a “wall” of electric treatment modules, physically contacting everything in its swath. By individually controlling the z-position of each electric treatment module, the “wall” may be changed in shape in accordance with what the imaging system perceives and in accordance with the transformation, to provide, effectively, spaces through which wanted plants can pass as the platform advances, and (conversely) wall portions that can come into physical contact with (or at least be proximal enough to for electrocution) unwanted plants.

[0042] In examples, the high density of the electric treatment modules in the array may be afforded by multiple rows of electric treatment modules. With a multi-row system, the individually-movable electric treatment modules may collectively form a negative three-dimensional space rather than merely a negative portion of a two-dimensional wall of electric treatment modules, which the negative spaces being for receiving wanted plants without physical contact thereby to avoid damaging the wanted plants, while collectively presenting this “wall” or - in the case of multiple rows - an electrocution region, for electrocuting unwanted plants.

[0043] It will be appreciated that, in examples, certain actual crop plants may be determined to be “unwanted plants” even though other crop plants of the same species may be determined to be “wanted plants”. That is, certain crop plants of a generally desirable species may be allocated for eradication, for the purpose of thinning the crop itself while, or instead of, actually weeding the crop by electrocuting actual weeds. Alternatives are possible.

[0044] In this description, “plant identification system” is a system that receives various types of plant data as input and is able to identify and distinguish between plant(s)-of-interest and plant(s)-of-no-interest such as unwanted plants (i.e. plants that are desired to be eradicated such as weeds), as well as to identify and distinguish between different kinds, maturities, and other attributes of plant(s)-of-interest, for the purpose of applying treatments, such as plant eradication treatments, in the same manner or in a different manner to different physical plants or clusters thereof.

[0045] In this description, “electric treatment” or “electrocution treatment” is an approach to eradicating or at least inhibiting growth of one or more targeted plants-of-interest (i.e. unwanted plants), using electricity. In this description, “mechanical treatment” is an approach to eradicating or at least inhibiting growth of one or more targeted plants-of-interest (i.e. unwanted plants), using mechanical means such as grasping a portion of the targeted plants with an end effector and pulling them out, cutting their stems with a blade of an end effector, scooping them out of soil with a scoop of an end effector, or a combination of these mechanical treatment approaches.

[0046] In examples, this disclosure is directed to an autonomous plant-treatment system, such as a robot comprising a carrier or platform, a control subsystem, an input subsystem, one or more manipulators for moving treatment modules with respect to the platform to locations within a three-dimensional region proximal to the platform, one or more power sources, and one or more electric treatment modules and / or one or more mechanical treatment modules.

[0047] The plants targeted by the system are, in examples, unwanted plants such as weeds or crops or any other plants. For example, crop plants may be targeted for eradication in order to thin a crop somewhat for the sake of overall health of the majority of plants of the crop. In an example, the system is intended for use in outdoor crop fields but can also be adapted for use in various other environments, such as within greenhouses, hoop houses, laboratories, grow houses, sets of containers, machines, or any other suitable locations.Control subsystem

[0048] In an example, the control subsystem may be an arrangement of devices (processors, memories, and communication modules) and software that directs and regulates the system’s actions. The control subsystem may interpret input signals from an input subsystem and external commands, process these input signals using pre-programmed instructions, eradication strategies and / or make on-the-fly decisions, and may send output signals and instructions to the robot’s actuators, electrical treatment system, carriers, and / or the input system. The system also includes a power supply for energy, and may include one or more feedback mechanisms to adjust performance based on the robot's actions taken. The control software executes processes that cause the robot to process data, make decisions, control its movements, and take other suitable actions. Together, components of the control subsystem form a control loop enabling the system to perform tasks autonomously or semi-autonomously, adapt to its environment, and handle situations the exact nature of which has not been specifically programmed in advance.

[0049] The control subsystem may be configured to index and assign an unique identifier to each object / plant in the field. In examples, it may gather identified objects from the input subsystem - an imaging subsystem, in embodiments, or a combination of an imaging subsystem and other input components - and associates them with their respective GPS (global positioning system) coordinates in the field, thereby generating a comprehensive map of a plurality of, or all of, the identified objects. It may operate to store, manage or / and update the unique identifier of each object / plant in the field as itgrows, going through various treatments, its ambient environments and plant characteristics change throughout time or any other suitable events.

[0050] As the input subsystem identifies a target, the control subsystem, which may include a primary processing subsystem, assigns the target to one or a plurality of actuators, electric treatment modules, applicators, soil sensors, effectors, and / or other suitable components. This may be done by translating the location of target in the frame of reference of the input subsystem to actuation locations and avoidance locations in a three-dimensional region proximal to the platform. Furthermore, the control subsystem may calculate the voltages and other parameters of the assigned treatment based on preprogrammed eradication strategies or form new strategies as the situations arise. That is, the control subsystem may be operable to determine a format of a time-limited electrocution treatment by selecting the format from a set of formats stored by the system based on one or more additional attribute detected and / or discerned by the input subsystem. Alternatively, or in some combination, the control subsystem may create the format of the time-limited electrocution treatment based on the one or more additional attribute.

[0051] An electrocution treatment may be informed by an eradication strategy: a set of optimal treatment parameters of the system in a particular treatment situation, such a weed eradication situation using electricity. The treatment parameters include, but are not limited to: the contact location on targeted plants / objects, distance between electrodes, parameters of applying electricity, discharging time, insertion path of manipulators, surface conductivity and form factor of the application electrodes, terraforming techniques, and other suitable parameters. A particular treatment situation means a situation with particular plant / object characteristics, ambient environmental characteristics, past treatments, and historic data. Different ones of these characteristics and treatment histories may be determinative of a different treatment situation. Eradication strategies stored in the control subsystem may be updated and / or created constantly based on the treatment and results of treatment.

[0052] Eradication strategies may incorporate empirical data or information derived from manual or automated observation and / or experiment to inform treatment methodologies or applications, such as electric eradication methods. The system may use past evidence and actual measurements to make informed decisions or predictions about plant treatment. For example, empirical data may include historical records of past plant moisture levels and their most effective treatments.

[0053] In some examples, the system can also be facilitated with the integration of statistical analysis, machine learning, deep learning, and computer vision, and other suitable Al (artificial intelligence) techniques or processes. These Al-driven methodologies enable the system to learn from its past operations, recognize patterns, make predictions, and visually or through other means to identify and classify plants / objects, plant growth stages, or ambient environment, and predict the optimal strategies of treatment and improve accuracy.

[0054] In some examples, the system may actively devise its own experiments by attempting various eradication strategies for the same or similar treatment situation and then test the efficacy or othercriteria of the treatment and store the optimal strategies. The strategies may include different parameters of the application electricity, insertion path, contact locations, terraforming techniques, discharging time, and distance between electrodes, surface conductivity and form factor of the application electrodes etc.

[0055] Alternatively, eradication strategies can also be derived from established scientific principles, mathematical models, or theoretical constructs. These guidelines assist in outlining the operational procedures of the system. For example, the optimal application electricity parameters for a treatment situation, preferably on a weed, can be derived using Joule's First Law: Q = I2Rt and other suitable relationships.

[0056] The eradication strategies stored in the control system can be any combination of the aforementioned methods.Controlled Artificial Lighting Subsystem

[0057] The input subsystem may be equipped with an artificial lighting subsystem. The artificial lighting subsystem incorporates one or more lights in forms including but not limited to lighting arrays, spotlights, LED panels, High Intensity Discharge (HID) lamps, Task Lights, or Flood lights. The lights may be capable of providing light with an intensity that matches or exceeds any external light sources that might affect the vision system (e.g. sunlight). This arrangement is designed to illuminate any surfaces of interest for applications including but not limited to imaging, object detection, object identification and object localization. Such an artificial lighting subsystem may be useful for providing a uniformity of lighting when ambient lighting from the sun cannot do so, such as when sunlight is occluded at times in a particular part of the field of view of an imaging device of the imaging subsystem by portions of a system platform or other components.

[0058] The placement and distribution of the artificial lights themselves can be valuable to ensure uniform and optimal photometric and radiometric illumination of the target area. Light sources may be strategically positioned in various configurations, including but not limited to overhead, side-mounted or integrated within structural elements to provide comprehensive coverage. The geometric arrangement of the lighting system is calculated using illumination modeling techniques to ensure uniform light distribution, minimize shadowing and reduce specular reflections. Some illumination modelling techniques include but are not limited to ray tracing; a rendering technique to simulate the distribution of light rays, and radiosity calculations to calculate light diffusion.

[0059] In some implementations, a curtain or skirt may be incorporated around an imaging subsystem and artificial lighting subsystem to deliberately inhibit or occlude any external light sources (e.g. sunlight). Such a skirt might be constructed from an opaque, light-blocking material such as rubber, polyurethane nylon or other light-blocking material or combination thereof. In some examples, slits may be cut into the skirt to provide some yield of the skirt material to reduce the instances of damage or harm to plants that an otherwise monolithic skirt might do. This configuration enhances theeffectiveness of the artificial lighting subsystem by reducing interference from external light sources, thereby providing more control over the lighting environment for the imaging subsystem.

[0060] The input subsystem may employ photodetectors to continuously or periodically monitor the ambient light intensity or the absence thereof. Based on the detected ambient light intensity, the artificial lighting subsystem may be caused to adjust its intensity and other parameters accordingly, including but not limited to power, intensity, degree of polarization, direction of polarization, luminous flux, color temperature, luminous intensity distribution, and wavelength (nm). Such an adjustment approach may tend towards optimal illumination while minimizing power consumption, thereby combining with and compensating for ambient light. In some examples, the artificial lighting subsystem may also adjust its parameters partially or wholly to accentuate one or multiple areas or objects of interest or to modify or enhance the image for computer vision purposes.

[0061] The artificial lighting subsystem described herein may be regulated by its own control subsystem (e.g. strobe circuit, programmable light emitters) or more generally the control subsystem serving the entire system. In either case, the control subsystem may manage the power to any light emitters in the system (e.g. LEDs) and synchronize the state of the light emitters to an imaging subsystem’s exposure time. To compensate for the high voltage, the light emitters may be operated at a low duty cycle for various reasons including but not limited to reducing overheating, extending the lifetime of the light emitters, etc.Imaging Model Training

[0062] Model training may be conducted for the plant detection system. Initially, as the system’s platform travels through a field, the imaging subsystem, which may include one or more imaging devices such as still image or video cameras, and other optical sensors, scans the field from a top-down perspective and identifies plants, other objects, and the overall arrangement situation. In some examples, the perception may be from the side rather than, or in addition to, top-down. The processing subsystem can receive images from the imaging subsystem and is configured to locate and identify a target in an image, such as differentiating between two plants, and measuring distance between two points in its perspective. In some examples, the perception may extend to include oblique views enabling the system to capture nuanced details and variations in plant morphology and spacing that might not be visible from other perspectives. The system may be arranged to incorporate multi-angle imaging, where multiple perspectives are utilized to create a 3D reconstruction of the field, enhancing depth perception and spatial understanding by generally gathering more data from a variety of perspectives.

[0063] To prepare the image data for further analysis, various preprocessing techniques may be performed. These techniques can include, but are not limited to, debayering, exposure control, resizing images, normalizing pixel values, white balance adjustments, cropping. Preprocessing may be useful to enhance the quality and uniformity of the input data, since this may directly impact the model’s ability to accurately detect and classify objects. In some examples, the imaging subsystem may includeone or more convolutional neural networks (CNN), processes of machine learning, computer vision, deep learning models, and other suitable artificial intelligence models.

[0064] In some implementations, sophisticated data augmentation techniques are used to address overfitting, particularly with limited datasets. Augmentation strategies may include randomized rotation, brightness, contrast, color temperature, two-dimensional blur, one-dimensional blur, gain, noise and any other suitable adjustments which improve the model’s ability to perform in diverse agricultural scenarios. By artificially increasing the diversity of the training data through such data augmentation, these techniques help the model adapt to the wide range of lighting conditions, plant orientations, and environmental variations encountered in real-world farming operations. Data augmentation techniques may be useful to provide model robustness and high precision for the image segmentation and contour masking steps, which can be valuable for targeting for electric weeding.

[0065] The plant detection model may be a supervised machine-learned model that describes a functional relationship between image data and predictions regarding the categorization of image data into masked instances. The model is typically parameterized, where a set of parameters captures the learned characteristics of the problem space. These parameter values are learned through training based on the labeled training data. The exact form of these parameters depends on the type of supervised machine learning technique used.

[0066] A wide variety of different types of supervised machine learning techniques may be used in place of a neural network including but not limited to Random Forests, Gradient Boosting Machines, Support Vector Machines (SVM), and other ensemble methods. The choice of model depends on the specific requirements of the plant detection system, including the complexity of the data, computational resources available, and the desired accuracy. Each type of model has its own strengths and weaknesses in terms of interpretability, training time, and performance on different types of data.

[0067] This plant detection system may utilize mathematical models for object detection and segmentation tasks. These models, often based on large-scale datasets including but not limited to COCO (Common Objects in Context) or ImageNet, have learned generic features from diverse visual data. These features can be transferred or fine-tuned for specific tasks like crop and weed detection. Transfer learning involves initializing the plant detection model with weights from the pretrained model and then fine-tuning these weights using the labeled image data specific to the agricultural environment. This process leverages the learned representations of general visual features, speeding up convergence and improving the overall performance of the model on the target task.

[0068] During training, hyperparameter tuning may play a role in optimizing the model’s architecture and training process. Hyperparameters such as learning rate, batch size, optimizer choice (e.g., SGD or stochastic gradient descent, Adam), and regularization techniques (e.g., dropout, weight decay) are adjusted to achieve the most appropriate trade-off between bias and variance in the model. For example, a higher learning rate may accelerate convergence but risk overshooting optimal weights, while a lower learning rate may converge slower but with more stable updates. Batch size affects the gradientestimation accuracy and training speed, with larger batches typically leading to faster convergence but higher memory consumption. Regularization techniques help prevent overfitting by penalizing overly complex models. The optimal configuration of these hyperparameters is typically determined through experimentation and validation on a separate validation dataset to ensure robust generalization performance.

[0069] In some implementations, the imaging subsystem may utilize one or more CNNs, machine learning processes, computer vision techniques, deep learning models, or other suitable Al approaches to process and analyze the image data.

[0070] In some examples, a self-supervised deep learning model for detecting crops and weeds may utilize the inherent structure in the data to learn useful features without requiring extensive labeled datasets. The system may automatically identify crops and weeds through visual pattern recognition in unlabeled data using techniques including but are not limited to contrastive learning, clustering, autoencoders or generative adversarial networks (GANs). By applying augmentations and transformations to create pseudo-labels, the model becomes adept at recognizing the subtle differences between plant species, ultimately improving the efficiency and accuracy of weed and wanted planted detection in real-world scenarios. This approach not only reduces the reliance on labor-intensive manual labeling but also enhances the model's ability to generalize across diverse environmental conditions and crop types.

[0071] In some examples, a semi-supervised deep learning model for detecting crops of wanted plants and weeds utilizes both labeled and unlabeled data to enhance its performance. The system may leverage a small amount of labeled data to guide the learning process while extracting patterns from a larger amount of unlabeled data using techniques including but are not limited to, consistency regularization, virtual adversarial training (VAT) and semi-supervised generative adversarial networks (SGANs). The combination of labeled and unlabeled data allows the system to generalize better across varying conditions and types of vegetation, reducing the need for extensive manual annotation and increasing the model's robustness in real-world agricultural applications.

[0072] FIG. 1A is an image, such as a frame of digital video, of a portion of a field in which a crop plant CP has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box R that is labelled / annotated “CROP 22 (SEMI-AUTO)” while the pixels corresponding to the crop plant itself have been changed in color as a contour mask.

[0073] FIG. IB is an image, such as a frame of digital video, of a portion of a field in which a weed plant WP has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box R that is labelled / annotated “WEED 8 (SEMI-AUTO)” while the pixels corresponding to the weed plant itself have been changed in color as a contour mask.

[0074] FIG. 1C is an image, such as a frame of digital video, of a portion of a field in which a weed plant WP has been identified and its location in the image delineated with a rectangular graphicaloverlay as a bounding box R that is labelled / annotated “WEED 273 (SEMI- AUTO)” while the pixels corresponding to the weed plant itself have been changed in color as a contour mask.

[0075] FIG. ID is an image, such as a frame of digital video, of a portion of a field in which a crop plant CP has been identified and its location in the image delineated with a rectangular graphical overlay as a bounding box R that is labelled / annotated “CROP 428 (SEMI-AUTO)” while the pixels corresponding to the crop plant itself have been changed in color as a contour mask.Annotation Process

[0076] The process described herein implements a binary approach to enhance the efficiency and scalability of plant detection. In an agricultural setting, in the field of view of the imaging subsystem, the plants within each image usually belong to diverse biological species, such as various weed and crop species. Treatments vary for different plants in commercial agriculture, for example, different weed species are controlled with specific herbicides or different parameters of electrical energy, while different crops receive tailored nutrients, agrochemicals, or pesticides.

[0077] In the binary approach, a plant detection system (e.g. computer vision model executed by the processing subsystem) may use classification methods to categorize detected plants as either plant-of-interest or plant-of-no-interest. This system may utilize machine learning or deep learning processes to analyze and extract visual features from plant images.

[0078] The processing subsystem may have been provided with several possible sources of labeled image data. In some examples, human labelers will utilize unlabeled data to generate labeled data. The subsystem may partially automate the process, by using semi-automatic annotation tools to detect objects and edges within the data being processed, minimizing manual labor and enhancing the efficiency and accuracy of data processing for the plant detection system.

[0079] The detection process undertaken by the processing subsystem will rely on a sophisticated model for plant-of-interest, including most growth stages and morphologies in the field. It serves as a baseline to compare against detected plants. In the weeding scenarios, plants conforming to this model will be classified as crops, while others are treated as weeds (e.g. weeds will be eradicated using electrical discharge). New weed types would still be detected and although they may have lower confidence level than a weed type included during training, they are likely to have higher confidence levels classified as a weed than as a crop plant. This crop-centric approach significantly reduces the complexity and demand of new training data and model training, focusing on creating a robust crop model rather than cataloging a wide array of weed types, offering adaptability and scalability to unseen weed species and fields.Masking

[0080] In some examples, a plant detection system may be used on a per image basis to physically locate distinct portions of image data that correspond to specific objects of interest (e.g., crops, weeds, obstacles). The plant detection system may utilize Convolutional Neural Networks, Recurrent Neural Networks, Encoder-Decoder architectures, Attention Mechanisms, Transformers, and any other fittingmodels. This task will produce contour masks, which provide information about the locations and sizes of each detected plant species or weed.

[0081] The process begins with the system analyzing the image data received from the imaging system to detect and differentiate individual instances of plants or weeds. For each detected instance, a contour mask is created, which provides pixel-level detection, offering advantages over object detection methods, which typically rely solely on bounding boxes that encapsulate an object’s approximate location. The contour mask technique provides the exact shape and boundary of each identified plant or weed, producing more accurate assessment of the plant’s area and morphological characteristics, including its area, shape, and boundary details.

[0082] The use of contour masks offers several advantages for farm operations, especially in areas of complexities for commercial farming where crops and weeds often interleave, occlude, or overlay each other. In such intricate scenarios, the accuracy provided by contour masking becomes valuable for conducting precise treatments using electric energy, as required by electric weeding systems. Electric weeding systems require significant precision to prevent wanted plants being contacted by electrodes, unlike spot spraying systems that may tolerate some mistreatment. The use of contour masks improves the system’s ability to apply selective treatments or interventions, such as delivering an electric shock to identified weeds, thereby improving the overall efficiency of the system and minimizing collateral damage to crops.

[0083] Additionally, the accurate assessment of plant morphology enabled by contour masks allows for better resource allocation for farmers who may use the data to tailor irrigation, fertilization and other agricultural measures, reducing water and fertilizer wastage and enhancing overall crop health and productivity.Input Subsystem

[0084] An input subsystem can be configured to collect a variety of data through passive and active collection methods. The input subsystem may collect plant data, such as plant morphological data, plant physiological data, physical data, biophysics data, plant ecological data, or any other suitable plant data. Additionally, the input subsystem may collect environmental data indicative of the ambient environment of the targeted plant / object, including but not limited to light, air quality and composition, atmospheric pressure, temperature, humidity, and precipitation.

[0085] The input subsystem may be configured to record measurements both while the system is in motion and when it is stationary within the designated geographic area. The input subsystem may include an imaging subsystem, a sensor subsystem, and a positioning module. In some aspects, as the platform travels through the field, the imaging subsystem, which may include cameras and / or various other optical sensors, scans the field from a top-down perspective and identifies plants, other objects, and their arrangement. In some examples, the perception may be from the side. The imaging subsystem may be configured to locate and identify a target in an image, such as differentiating between two plants, and measuring the distance between two points in its perspective.

[0086] In some examples, plant data are preferably collected as optical data using cameras such as stereo cameras, multispectral cameras, CCD cameras, hyperspectral imaging systems, LIDAR systems, dynamometers, IR cameras, thermal cameras, single-lens cameras, monocular cameras, projected light imaging systems, scanning imaging systems, time-of-flight systems, optical sensors, imaging sensors, and other suitable sensors. Alternatively, plant data can be reflectance data, electric data, x-ray data, ultrasonic data, tactile data, thermal data, chemical data, or other suitable data. Plant data can also be collected using a variety of other sensors, such as spectrophotometers, LiDAR sensors, multispectral cameras, voltmeters, ammeters, multimeters, electric field sensors, Electrical Impedance Tomography (EIT) sensors, x-ray detectors, x-ray cameras, x-ray spectrometers, ultrasonic sensors and transducers, tactile sensors, force sensors, pressure sensors, infrared thermometers, thermal cameras, thermocouples, Resistance Temperature Detectors (RTDs), thermistors, gas sensors, ion-selective electrodes, pH sensors, conductivity sensors, mass spectrometers, accelerometers for motion data, humidity sensors, microphones, acoustic sensors, touch sensors, electromagnetic sensors, or other suitable sensors.

[0087] In some configurations, the system may incorporate multiple variations of high-resolution (zoomed-in) and low-resolution (zoomed-out) cameras. The high-resolution cameras are configured to capture detailed plant information, which is subsequently processed by the plant identification system. These high-resolution cameras provide precise data on individual plants, including morphological and phenotypic characteristics. Conversely, the low-resolution cameras are designed to capture broader, holistic plant data, which is utilized for route mapping, robot localization, and navigational purposes. The low-resolution cameras offer a wider field of view, enabling the system to effectively monitor larger sections of the agricultural environment and facilitate the overall movement and positioning of the robotic platform.

[0088] In some examples, active collection methods may include applying stimuli to trigger responses from a plant / object, such as electric stimuli, light stimuli, tactile stimuli, and thermal stimuli. Ambient environmental data may also include soil compaction, soil color, soil temperature, soil moisture, soil texture, soil pH, soil particle size, soil content (nutrients, chemicals, and other soil sampling data), soil drainage, wind speed, gas emissions, or any other suitable data.

[0089] In some examples, the data collected may also include information regarding other plants / objects in the vicinity of the targeted plant / object, such as pest information (including disease, insects, animals, etc.) in the environment, weed height and pressure in the vicinity, cover crops grown nearby, and other suitable data. Ambient environment sensors may include a camera, a temperature sensor, an ambient light sensor, a photodetector, a humidity sensor, a pressure sensor, an electric conductivity capacitive sensor for in situ soil analysis, a pH sensor, an ion meter, a conductivity sensor, a moisture sensor, a seismograph, a wind velocity sensor, or any other suitable sensor.

[0090] In some examples, the input subsystem utilizing cameras, plant sensors, and other sensors can determine a wide variety of plant characteristics. These characteristics may include leaf properties such as size, area, temperature, mass index, color, color distribution, weight, texture, edge roughness, andfuzziness. Whole plant characteristics may include size, color, color distribution, estimated height, volume, mass, water content, density, lightwave absorption or reflection, and sugar concentration and its distribution across the plant. Information on the plant's crown diameter, stand count, leaf uniformity, and plant conductivity may also be determined. Fruit-related characteristics may include size, color, sugar concentration, water content, quantity, distribution, skin thickness, weight, density, and smell. For stems, characteristics such as woodiness, chemical composition, width, and elasticity may be measured. Information regarding grains, such as quantity, moisture, and compound content, may also be gathered. Additional factors like the relative position of plant components, root distribution, aeration, depth, thickness, or other suitable plant characteristics may also be gauged.

[0091] Plant or ambient environmental characteristics may also be extracted from the data collected. For instance, plant moisture levels may be estimated with the temperature difference between the ambient environment and the plant. Soil moisture levels may be estimated by measuring the reflectivity of soil in the targeted area. Weed pressure in the area, nearby weed species, cover crop distribution, cover crop species, disease, and pest exposure, and plant cluster and overlay situations may be extracted using computer vision techniques. Soil content and chemistry may be estimated using electroconductivity capacitive sensors for in situ soil analysis or any other suitable ambient environmental characteristics.Object Indexing

[0092] The system can function to autonomously identify a plant / object within the geographic area. In some examples, the system identifies the plant based on measurements from the imaging subsystem, or it can identify the plant in any other suitable method using other sensors installed in the input subsystem. An input subsystem with various sensors included may be configured to receive, generate, and / or derive sensor data from any number of sensors as a vehicle travels through a geographical area. In some examples, the processing subsystem or control subsystem generally will create a mapping of the field and assign value and indexes to each plant, and each plant is identified and indexed using a unique identifier. The unique identifier may be based on its GPS location, plant species, plant structures, sizes, plant morphological, physical or physiological, biophysics, and plant ecological characteristics, the plant / object ambient environment characteristics, and / or previous treatments.

[0093] The input subsystem preferably sends the measurement to the control system for processing and plant identification, however, in an alternative configuration, the input subsystem possesses the capability to process the measurement independently to identify the plant.

[0094] The input subsystem may be configured to sense, detect, analyze, store, and / or communicate data associated with one or more objects / plants. The imaging subsystem, which may be considered a part of the input subsystem, may be configured to detect and identify one or more objects / plants on its path. It can also be configured to detect subsets of one or more objects / plants ex. a stem, a leaf, a flower, apical meristem, branches, fruits, a root system, or any portion of a plant, and may detect one or more objects adjacent to the objects / plants, such as pests, weeds, soils etc.

[0095] All information stored in association with the plant can be retrieved based on the unique identifier. In some example. Plants can also be identified based on their species or genus. The unique identifier can also include a geographic location, such as GPS coordinates; plant features, such as size, growth stages, species etc.; other objects exist on or nearby the targeted plants / objects, such as pests, crops, disease pattern; previous treatment, or any other suitable traits.

[0096] The input subsystem may consist mainly of the imaging module / subsystem and the positioning module. The positioning module may supply data configured to locate an object (preferably a plant) relative to carrier, to actuator to a reference coordinate system (GPS) or to any other object. It may include radar, sonar, ultrasonic sensor, gyroscope, accelerometers, odometry, or any other suitable sensors. Position information can also be carried by QR (quick response) code or barcode.

[0097] The subsystem would record the GPS location of the plant, growth stages, the treatment conducted, various plant data, and ambient environmental data that are appropriate. The data gathered by the input subsystem can be utilized to build a chronological model of a plant's physiological conditions and growth stages, associated with a unique identifier.

[0098] In some examples, the input subsystem determines the position of an actuator or relative to an object, preferably a plant. One or multiple sensors in the input subsystem may be distributed among any part of the carrier in any combination. The input subsystem can then determine the position of an actuator in relation to an object / plant, and will be configured to identify that object / plant, target it, and / or perform an action or apply a treatment autonomously.Historic Data

[0099] The subsystem can additionally function to leverage historical data. Historical data can include historical treatment data, historical weather data for the geographic area of the targeted field, historical planting data (including planting depth, plant genetics and epigenetics), historic weed pressure, weed height, and weed species, historic growth stages and plant size / weight at particular time, historic plant characteristics, ambient environmental characteristics, plant genetics or any other suitable historical data.

[0100] Treatments can include electric weeding, laser weeding, hand weeding, cultivation, burning, plant hormone application, salinity control, fertilizer application, water application, drought, fungicide application, pesticide application, other chemical application, fruit harvesting, disease exposure, insect infestation exposure, or any other suitable plant treatment. That is, a given treatment may generally be one of: an eradication treatment (in the case of, for example, weeding or burning), an improvement treatment (in the case of, for example, water, fertilizer or pesticide treatment), or a harvesting treatment (in the case of, for example, cultivation).Actuators

[0101] Supported by the platform is an array of multiple actuators, each supporting and enabling linear z-direction motion of a respective electric treatment module. In an example, each of the actuators is fixed in the x-y (i.e., left-right and front-back) directions with respect to the platform during a particulartreatment, thus fixing the respective electric treatment module in the x-y directions, but is operable to cause the respective electric treatment module that it supports to be moved in the z direction (up-down). This z-direction adjustability is to enable the electric treatment module to be controlled to assume an actuation location - a z-position of an electrode of the electric treatment module that corresponds physically with an unwanted plant so that the electric treatment module can then electrocute the unwanted plant - or an avoidance location - a z-position of the electrode that does not correspond physically with a wanted plant so as not to physically interfere with the wanted plant. It will be appreciated that the actual actuation location will vary based on a number of factors, such as the height of the unwanted plant to be electrocuted. Different unwanted plants may have different heights, and thus a given electric treatment module may be positioned at one z-position to physically contact one unwanted plant and may be positioned at another z-location to physically contact another unwanted plant. Furthermore, the actual avoidance location for a given wanted plant may vary. Different wanted plants may have different heights, and thus a given electric treatment module may be positioned at one z-position to just physically avoid contact with one wanted plant and may be positioned at another z-location to just physically avoid contact with another wanted plant. Alternatively or in some combination, the system may be configured such that the actual avoidance location of an electric treatment module may be fixed such that the electric treatment module is caused to be always withdrawn to only a highest z-position when it is required to move to its avoidance location, such that the avoidance location does not depend on the height of the wanted plant.

[0102] Various motors, transmissions, and position feedback sensors may be provided in each actuator, controllable by the control subsystem responsive to the input subsystem, to efficiently and quickly raise and lower the respective electric treatment module repeatedly to accurate actuation locations and avoidance locations as the field is traversed by the platform for weeding and / or crop thinning, thereby to physically avoid wanted plants and to physically contact unwanted plants for delivering an electrical treatment. See, for example, FIG. 19, which is a diagram showing a representation of a crop with wanted and unwanted plants that an arrangement of treatment modules, such as tines, may address for treatment.

[0103] The structure of the actuators and their respective electric treatment modules may be modular and scalable. In particular, a system for weeding may be equipped to support different numbers and spatial configurations of electric treatment modules in its array, such that a user or systems integrator may easily equip a given system appropriately for an application, based on different field sizes, crop densities, numbers of beds (a bed being, for the purpose of this description, a row of soil where crops are systematically planted in straight lines with competing weeds), and overall cost considerations (more electric treatment modules may correspond generally to high overall cost of the system, so users may wish to control costs by purchasing systems with a less-dense array of electric treatment modules but that nevertheless provides a satisfactory enough weeding / thinning functionality for the application). The electric treatment modules may be arranged in a desired pattern in the array.

[0104] A given array of electric treatment modules may be supported on a frame that itself can be automatically or manually raised or lowered with respect to the platform that supports it, to cause all of the electric treatment modules to assume a home position that is higher or lower with respect to the crop. Such height adjustment would typically be done when the system is not operating, to enable the system to be conformed to different heights of crop before treatment begins. That is, a newer planting may require the frame supporting the array to be lowered so that the electric treatment modules can reach down far enough to physically contact shorter unwanted plants, but a more mature planting - later in the season perhaps - may require the frame supporting the array to be raised to that the electric treatment modules are not, in their home positions, being dragged through the tops of all of the taller crop even though at times they will be required to be in avoidance locations. Variations are possible.

[0105] Various configurations can be tailored to ensure effective weeding operation across diverse conditions, such as varying soil types, crop arrangements, crop types, growth stages of crops and weeds, and weed densities. The spacing, arrangement, combinations, angle, and orientation of the electric treatment modules (or “tines”) will be carefully adjusted to achieve maximum clearance, reduce the risk of damage to crops, and enhance the overall efficiency of the system. Additionally, modular or adjustable tine arrangements may be incorporated to allow for quick adaptation to changing field requirements, ensuring versatility and precision in addressing unique farming challenges. In some examples, tines may be designed in different shapes, cross sectional areas, swath width, or form factors to suit different use scenarios. This is also shown in FIG. 19.

[0106] In certain embodiments, tines need to be densely packed to achieve the level of precision required for effective weeding or crop thinning. However, due to physical limitations and insufficient clearance between individual tines, it may not be feasible to arrange tines in a single line while providing the desired density. To address this challenge, multiple tines may be arranged in a multi-layered configuration, where each layer is carefully designed to optimize spatial utilization and ensure sufficient clearance between tines.

[0107] This layered arrangement not only resolves the physical clearance constraints but also enhances the overall fineness and precision of the weeding process. By enabling the tines to be positioned closer together across multiple layers, the system can deliver a higher degree of accuracy in targeting weeds, ensuring comprehensive treatment of the covered areas. Additionally, the multi-layer design can be tailored to adapt to different field conditions, crop layouts, and weed densities, providing flexibility and effectiveness in a variety of agricultural scenarios.

[0108] Furthermore, the multi-layered structure can be tailored to accommodate varying crop heights, weed densities, and field conditions, ensuring versatility and adaptability across diverse agricultural scenarios. This approach not only maximizes weeding accuracy but also improves overall efficiency in challenging and high-density farming environments.

[0109] In some examples, one or multiple cameras of the input subsystem may be assigned to each layer of tines.

[0110] In some examples, the camera vision may have an assigned targeting field of view to cover the required distance of a given number of crop beds

[0111] In some examples, each tine may be assigned a horizontal strip of the crop bed where it is solely responsible for zapping the weeds and avoiding the crops. For example, in a 50cm wide crop working area where there are 10 tines, each tine will be responsible for the area 2.5cm to each side of itself as the robot moves down the crop bed.

[0112] In some examples, multiple cameras may be mounted alongside multiple layers of tines. In some examples, each layer of tines may be assigned one group of cameras. Other arrangements are possible to achieve the necessary vision-manipulation synchronizations.

[0113] In some embodiments, by default, the tines will maintain a fully-extended position where they hover above the ground as the robot moves forward.

[0114] In some embodiments, the exact position of the tines will be recorded by the system at all times.

[0115] In some embodiments, when the camera vision system detects that a crop is present, the position of each pixel on the crop will be reported to the control subsystem in order to determine which tine to actuate - to put in an avoidance location or in an actuation location - and at which time. In examples, the ‘x’ value of the pixel may be used to index which tine is meant to be actuated. For example, if a 1920 x 1080 (width x height) pixel camera is used with a 10-tine implementation, pixels (0, y) to (191, y) will represent crops that must be avoided by the first tine. Similarly, pixels (192, y) to (383, y) will represent crops that must be avoided by the second tine, and so forth until pixels (1728, y) to (1919, y) represent the tenth tine. Other ways of Pixel-to-Tine Mapping or pixels-to-tine mapping are possible to achieve mapping.

[0116] The ‘y’ value of the pixel will be used to determine how much delay before the indexed tine must actuate, as well as the time before it is allowed to come back down to fully extended. For example, if a 1920 x 1080 (width x height) pixel camera reports the lowest pixel on a given crop as (x, 500), it can be determined this point is 580 pixels from the bottom of the frame. For example, the highest pixel on the same crop is (x, 400), thus being 680 pixels from the bottom of the frame.

[0117] Before the robot is to identify and eliminate crops, it will first be calibrated to find the effective distance per pixel in the image . For example, if a 1920 x 1080 resolution camera is covering an effective width of 1.08m, it can be found that the distance per pixel is Imm / pixel. This means that the crop's lowest point is 580mm, or 58cm from the bottom of the frame, while the highest point is 680mm, or 68cm from the bottom of the frame. This, in conjunction with the known distance from the tines to the bottom of the frame, the processing subsystem can provide the total distance from both the crop’s lowest and highest points, to the indexed tine that is responsible for avoiding it.

[0118] The time taken before reaching either of these points can be found through a function of the robot’s horizontal velocity and the total distance from the point to the indexed tine.

[0119] The tine will be pulled up before reaching the closest point, and lower down after fully clearing the furthest point.

[0120] In examples, a protective enclosure surrounds each tine which will be water- and dust-resistant. This will protect the tine from all sorts of weather as well as dust that will be kicked up from the operation of the system in the field as well as from wind, external factors etc.

[0121] In some examples, the device used to carry out the weed treatment may have a different delivery system, method of weeding, and form factor, utilizing technologies such as a laser, sprayer, mechanical gripper, and mechanical uprooting device, instead of electrocution.Electric Treatment Modules

[0122] According to an example, each electric treatment module of the system includes an apparatus equipped with electrodes designed for applications including but not limited to, killing a plant or attenuating plant growth. These electrodes are caused to be positioned to deliver controlled electrical discharges upon contact with a weed. When the electrode comes into contact with a weed, current passes through the plant. A high voltage discharge through the weed’s stem can lead to the explosion of cells within the plant. In instances of sufficient current strength, the electrical energy can traverse through the entire plant forming an electric circuit, effectively killing it at the root. Research on electric weeding has shown that, with the adjustment of electrical energy parameters, it can be effective against weeds at all growth stages. When eradicating or attenuating weeds with electrical energy, it is generally similarly effective regardless of where on the plant the application electrode delivers the electric energy, contrary to laser weeding, which requires targeting the apical meristem, or mechanical weeding, which requires picking at the stems to pull out the roots. When eradicating or attenuating weeds with electrical energy, it requires only a short contact period (a fraction of a second) for effective treatment. The application electrode needs to be in contact with each weed for a negligible amount of time, making it an efficient method for commercial weeding.

[0123] In an example, each electric treatment module contains one or more electrical energy processing units that convert input electricity from the power source to the required, adaptable output parameters, including voltage, current intensity, waveform and frequency, designed for optimized weeding efficiency. This setup includes application electrodes for precise and targeted energy delivery to plants, enabling effective treatment, and earth electrodes for safe grounding, ensuring operational safety and efficiency. The module is integrated with a sophisticated control subsystem, which orchestrates the adjustment of these parameters based on real-time data and predetermined strategies for effective weed eradication. This control subsystem manages the timing and intensity of electrical discharges, ensuring precise and efficient treatment while minimizing the impact on desired crops or surroundings.

[0124] The electrical energy processing unit comprises a converter configured to receive unprocessed electrical energy from an electrical energy supply and convert the unprocessed electrical energy to processed electrical energy. The said processed electrical energy is output to an applicator unit. The converters may be configured to transmit processed electrical energy between an applicator electrode of an applicator unit and an earth electrode of an earth unit. In some examples, the converter may be configured to transmit the processed electrical energy circuit.

[0125] The module operates in conjunction with the control subsystem and imaging subsystem to autonomously identify and selectively treat weeds in agricultural fields. It is controlled by the movement subsystem (in some embodiments, a drive subsystem that drives wheels or tracks through a crop) which is itself controlled by the control subsystem thereby to move the electric treatment module(s) relative to the platform supporting the modules to actuation locations in coordination with moving of the platform by the movement subsystem. The route of each module and the control of the movement system is based on desired location data from the route mapping process, computed by the control subsystem.

[0126] The control subsystem will determine one or more relevant eradication strategies for a targeted plant / object using the data supplied by the input subsystem, for example from the imaging subsystem, about the targeted plant / object. The control subsystem also controls the discharging time of the electrical energy and converts the determined eradication strategies to specific instructions, and communicates the said instruction to electrical energy processing units and manipulators to execute electrical treatments.

[0127] In some examples, the electric treatment module is in the form of one or more electrode bars in the form of hollow rods and are made of conductive materials including but not limited to copper, copper alloy, or the like, that can array and discharge electrical current. The electrode bars are positioned at a height above the ground where they can come in contact with the weeds. The electrode bar is positioned so that they contact the weeds and not the desired crop.

[0128] In some examples, a Faraday enclosure includes multiple layers to optimize blocking of electromagnetic radiation, with the first layer made of materials including but not limited to, tin, permalloy, or supermalloy for lower frequencies, and the second layer made of materials including but not limited to, copper or aluminum. A Faraday enclosure is electrically connected to the ground to safely transmit leaked electrical energy back to the transmission circuit. An earth electrode may be used to physically contact the ground, providing an electrically conductive connection. The treatment mouth of a Faraday enclosure refers to the opening where plants of interest are directed for electrical treatment. The enclosure is designed to enclose the applicator and return electrodes with minimal gaps to block electromagnetic radiation effectively.

[0129] In some examples, the electrical energy processing unit is configured to close-loop control and regulate the specified parameters of electricity being applied to plants / objects according to the instructions from the control system. The specified parameters include but are not limited to voltage, current, power, frequency, phase shifts, phase, waveform, and may involve various forms such as AC (alternating current), DC (direct current), pulsed, and other suitable parameters. The electrical energy processing unit includes various sensors to monitor the output parameters of electricity, as well as various windings, circuitries and associated conversion devices to modify the input parameters according to the instructions from the control system.

[0130] The form factors of electrodes can vary depending on the application's specific weeding efficiency requirements. In some examples, the route mapping system can adapt to different electrode shapes; for example, if wider electrodes are used that can treat multiple weeds along their path, the route mapping process would consider the width corresponding to the number of weeds targeted per pass.

[0131] FIG. 2 is a schematic diagram showing a representation of an electric treatment module (or “applicator”) 40 supported by a platform (not shown) and having an electric applicator 44 extending from a body 42 and in contact with a weed 100 growing out of ground G. These components form an electric circuit 70 with an earth unit 60, also supported by the platform, which has an earth electrode 64 extending from a body 62.Eradication Strategies

[0132] Some eradication strategies may include but are not limited to the following.

[0133] Terraforming strategies: The efficacy of electric weeding may be impaired if the targeted plants and / or ambient soils have low moisture levels. When the input subsystem senses that the soil, the targeted plants or the ambient environment are dry, it implements higher voltage electric weeding or may spray water either in the soil or on the plants before treatment to ensure efficacy of electric weeding, and to ensure the electrical energy is just enough to kill the weeds / plants / other objects without physical damaging the system or posing risks to the human users nearby.

[0134] The subsystem can change the distance between application electrodes and earth electrodes. It can be determined based on the soil content characteristics and the species, growth stages, and size of the targeted plant / object.

[0135] In some cases, different plant species require tailored parameters of electricity, ex. voltages, frequencies, phase shifts, waveforms, discharging time to eradicate effectively and optimally. In some cases, different growth stages of plants require tailored parameters of electricity, ex. voltages, frequencies, phase shifts, waveforms, AC / DC and other suitable properties, to eradicate or to treat effectively and optimally. For instance, sub-5000 voltage DC electricity can effectively eradicate any 2-leaf and 4-leaf stage weeds in general in less than 0.1 seconds. Lambsquarter at maturity requires >15000 volts, 20 kHz AC electricity with minimum 0.5 seconds discharging to treat it effectively.

[0136] During treatment, the control subsystem and / or processing subsystem thereof will identify plant species and growth stages and adjust or keep the default parameters of electricity to ensure optimum efficacy of treatment or eradication to each individual plant or object. In some cases, certain plant species, such as late stage weeds with adventitious root systems, require multiple electric treatments at various locations (such as zapping directly above all roots) of the plant to ensure eradication.

[0137] In some examples, to maximize the efficacy of electric weeding, the actuators (in this case, application electrodes) may be controlled to target different locations or organs of the plant (any below the soil surface portion in the direction perpendicular and parallel to the surface line, and above the surface portion in direction perpendicular and parallel to the surface line) depending on the previoustreatment location, the overlay and cluster situations with non-targeted plants or objects, and the orientation of the plant growth.

[0138] An overlay situation occurs where targeting plants grow underneath one or multiple layers of non-targeting plants or objects, with only small portions to be seen from the imaging system perspective, or where non-targeting plants or objects grown underneath targeting plants. A cluster situation occurs where targeting plants and individual non-targeting plants or objects grow close together from the imaging subsystem perspective.

[0139] In some examples, the control subsystem may adjust the distance between an earth electrode and an application electrode to maximize the efficacy and the accuracy of treatment. The system can perform additional eradication strategies (ex. spraying other chemicals) in conjunction with electric weeding.

[0140] The system includes safety controls, for example upon detection of electric leakage above a certain threshold, the control subsystem may temporarily cease electrical weeding, and wait for further instruction.Electrocution, or “Zapping” Location Determination

[0141] An eradication map can be employed by an agricultural machinery to facilitate the targeted treatment of plants within a specified field. In various examples, the eradication map may comprise a 2 or 3-dimensional data matrix, a collection of image masks and shapes highlighting each instance of a plant, or a set of plant coordinates, wherein each coordinate may correspond to a specific portion of one or multiple plants. Each map element within the eradication map is associated with a treatment area, defined as either a coordinate point or a coordinate point range, within a referenced image. Additionally, each map element encompasses information pertaining to identified objects within the associated treatment areas, such as classification as a crop or weed, dimensional attributes, color characteristics and so on. Consequently, with information derived from the eradication map, the actuators of an agricultural machinery will conduct field treatment in accordance with predetermined eradication strategies.

[0142] For instance, a control subsystem is configured to receive the information eradication map from the processing subsystem and generate corresponding control signals for the actuators and the applicators themselves. These control signals are subsequently transmitted to the actuators of the agricultural machinery system to execute corresponding eradication strategies. In some examples, the control signals are timed such that an actuator is activated when it is positioned directly above the identified object. For example, the eradication map may be represented as a matrix data structure, wherein each map element is associated with a potential weed location. In this context, a map element of the eradication map indicates that the associated treatment area within the referenced image contains pixel data indicative of a weed. Consequently, the control subsystem generates control signals for the actuator that is linked to the specific map element, ensuring that the actuators administer treatment to the targeted weed as the agricultural machinery traverses the field.

[0143] In some examples, the actuators must be controlled to make contact with the plant, for example, if the actuator is an electrical discharge device / application electrode, it may make contact with targeted plant to deliver electrical discharge to eradicate the plant. In this case, there is a process that determines the optimal sections of the eradication map and the individual plant markers (in some examples, a range of coordinates for each plant / group of plant instances to be within) to move the electrical discharge device to, in order to eliminate the weed.

[0144] The plant identification markers generated by the imaging system are used to identify one or multiple points which are evenly spaced across the whole marker, and these points are located near the center of the plant. In some examples, this process will ensure that the targeted point on the plant is near the center to allow for possible slight positioning error while still eliminating the plant. The program works by clustering points that are close to each other into groups, creating effective electric treatment clusters. The program takes a parameter for a maximum distance value between two points of a cluster, and also a minimum number of points in a cluster. It begins by visiting data points and identifying their neighboring points. Then, the distance between neighboring points is compared to the maximum distance, determining whether or not these two points are a part of a cluster. After one cluster is formed and no more neighbouring points remain, the next unvisited point is visited and marked as a different cluster, and the process repeats as its neighbouring points are visited as well. This will continue until all points have been visited and the proper point clusters, representing leaves or individual plants, have been generated. The purpose of this is to eliminate any parts of the plant marker that may be outside of the main plant area, where the leaves / stem are. Next, the centroid of the targeted clusters is calculated mathematically and the targeted point is set to the centroid. This is how the targeted point is determined in some examples.

[0145] This approach also ensures that in the case where there are multiple weed instances in particularly large plant identification markers, most of them will be targeted. In conventional commercial agriculture, the majority of weeds within a single workstation of agricultural machinery are typically at the same growth stage and exhibit similar two-dimensional areas from a top-down perspective, which means the weed centers of mass are roughly of similar distance from each other. Electric weeding is generally equally effective at eradicating weeds regardless of where the electrode contacts the plant. Therefore, this method can efficiently approximate the locations of each weed within a large plant identification marker without necessitating the computational resources and time to individually identify each weed within the marker. In some examples, the number of treatment locations identified can be set to change based on the size of the plant identification markers.

[0146] This collection of treatment locations, in some cases coordinates, is then used as the input into an electrical applicator movement control subsystem, which will use the treatment locations outputted by the system to determine the motion of the electrical applicator. More specifically, this process transforms the targeted plant markers in the eradication map, received from the data input and plant identification system, into a set of precise locations in the three-dimensional region proximal to theplatform to which the electrical applicator must be moved to in order to eradicate all target plants within the given time window (corresponding to the information captured at a particular time by the imaging subsystem, for example). This set of precise locations will then be inputted into the route mapping system which will determine the optimal route for the electrical applicator to follow to reach all locations. Then, the movement control subsystem will control the actuators to move the applicator.

[0147] FIGS 3 A, 3B which are images of portions of fields, and FIG. 4 which is a graphical representation of two overlapping portions of a weed, provide an example of the possible contact points P that the electrical applicator must reach in order to eliminate the targeted plant. In some examples, these points would be mathematically determined for every plant location marker R, as an example. In some cases, more contact points may be assigned to larger plant location markers and conversely, less contact points may be assigned to smaller plant location markers, as seen in the figures.Plant Location Tracking and Depth

[0148] The disclosed system pertains to a plant location tracking method employed by an agricultural machinery, facilitating ongoing monitoring and treatment of targeted plants within a designated treatment area. Utilizing advanced tracking processes, the system ensures precise delivery of electrical discharges to the identified weeds to optimize the effectiveness and efficiency of the weeding process, by avoiding duplication or omission of treatment.

[0149] Continuous tracking is performed, and as the platform supporting the agricultural machinery traverses the field, new video frames are continuously captured by the input system. The tracking program updates the position of each identified weed in real-time, ensuring accurate tracking of their movement. In some examples, the system integrates visual data from an imaging subsystem with examples including but not limited to, IR sensors, cameras, hyperspectral cameras, multi-spectral cameras, and depth data from a depth sensor with examples including but not limited to, LiDAR, sonar RGB-D (red, green, blue, depth) cameras, stereo cameras, Time of Flight sensors, and radar. The process described herein allows the system to maintain precise tracking even when weeds move or change appearance due to a change in environment factors such as lighting or shadows.

[0150] The input subsystem, in some examples, includes components arranged at the front of the platform - at least, frontward of the electric treatment modules - captures initial frames of the crop rows, and the processing subsystem processes them as described herein to identify individual plants-of-interest which may include weeds, crops or any plant. It then detects the location of a target plant based on this image and calculates the location of the actuator, which is both longitudinally offset from and laterally aligned with the plant's position. The control subsystem of the system subsequently drives an actuator to move with respect to the platform, and in coordination with the moving platform, to align with the plant’s location. This electric treatment module is housed behind the artificial lighting module. The method involves tracking the plant’s location relative to the actuator’s reference position. Once the module is aligned with the plant location, the electric treatment module is actuated.

[0151] For each detected plant-of-interest, a program will be initiated, which is used to continuously estimate and predict the positions of objects over time, based on sequential observations or measurements, including but not limited to examples such as SORT (Simple Online and Realtime Tracking), Kalman Filters, or deep learning based trackers.

[0152] In some examples, the method begins with capturing a first image of the crop rows below an artificial lighting system arranged near the front of the platform. The location of a first target plant is detected based on this first image, and a first actuator location with respect to the plant is calculated, which is longitudinally offset from and laterally aligned with the plant's position. The machine then drives the actuator to move to laterally align with the plant’s location. A second image of the crop rows is then recorded. This second image is used to verily the position of the actuator relative to the plant location and to calculate any necessary longitudinal and lateral offsets between the first actuator location and the actuators' reference positions. These offsets are then applied to correct the actuator's position for more accurate alignment and operation in subsequent actions.

[0153] This process is repeated for subsequent target plants. For each new plant, the machine records another image, detects the plant's location, and calculates a new actuator location. This new location is corrected using the previously calculated offsets.

[0154] The system may store and deploy strategies to manage temporary occlusions, where weeds might be blocked from view by other plants or obstacles. The utilization of predictive models including but not limited to Kalman Filters, Particle Filters, Support Vector Machines (SVMs), Neural Networks, and Bayesian Networks, can estimate the position of weeds during these occlusions, enabling the continuation of tracking. These models may utilize historical data or real time measurements to predict the positions of occluded weeds by utilizing data learned from pre-trained models, learning process and probabilistic models.

[0155] When occluded weeds reappear, the system may re-establish tracking by matching the reappeared weeds with their predicted positions by using techniques including but not limited to Re-Identification process and Correction Models. These techniques leverage extraction methods (e.g. SIFT, SURF) and machine learning classifiers to match reappeared weeds with their predicted positions based on visual cues such as shape, color, size, etc.

[0156] Depth information in the system’s field of view is collected, using depth sensors with examples including but not limited to, LiDAR, sonar RGB-D cameras, stereo cameras, Time of Flight sensors, and radar. In some examples, a multiple-camera system itself is used as the depth sensor to capture depth data in the three-dimensional region of interest. Such a multiple-camera system captures depth information by using two or more cameras positioned at a fixed baseline distance apart, mimicking human binocular vision. When the cameras simultaneously capture an image of the same scene from slightly different perspectives, objects / pixels appear displaced between the left and right images and may be referred to as disparity. The system first rectifies these images to align corresponding points along the same horizontal scanline, ensuring accurate disparity calculations. Using feature-matchingalgorithms like Block Matching (BM), Semi-Global Matching (SGM), or deep-leaming-based methods, the system identifies corresponding pixels in both images and calculates the disparity, which is the pixel offset between the left and right images. Depth information is derived using triangulation, where the disparity, camera focal length, and baseline distance are used. Closer objects / pixels exhibit greater disparity (larger pixel shifts), whereas farther objects / pixels have smaller disparities. The computed depth values for each pixel generate a dense depth map, allowing the multiple-camera system to reconstruct a 3D representation of the scene. In some situations, the cameras / sensors that collect the imaging information of plants, also collect the depth information in similar fashions described above. The depth information is collected and mapped in the same resolution as the imaging system, where each depth pixel is re-projected onto the 2D image plane pixel by pixel to form 3D mapping in the imaging system’s field of view. In some situations, the cameras / sensors that collect the imaging information of plants, are installed in different locations from the depth sensor. Spatial calibrations will be performed to integrate the reference frames to form a 3D model for the area of interest.

[0157] In some examples, a laser system may be used to collect depth data by measuring distance by calculating the time it takes for a light pulse (usually infrared laser or LED) to travel to an object and reflect back to the sensor. Such a laser system emits a modulated light signal toward the target, and a photodetector captures the reflected light. By measuring the round-trip time delay (At) between emission and reception, the sensor computes the distance between the sensor and the objects / pixels in regions of interest.

[0158] In some examples, the depth sensor has the same resolution as the imaging sensor that conducts plant detections. A fusion algorithm may be used to convert depth pixels to 3D world coordinates and re-projects them onto the 2D image plane provided by the imaging system in pixel by pixel fashion to form 3D mapping. In such a system, this conversion may be done in real time during the field operations.

[0159] In some examples, the depth sensor may have lower or higher resolutions than the imaging sensor used for plant detections. The depth sensor will provide critical depth data of one or more pixels within the object regions (detected crops, weeds, ground, obstacles etc.) necessary to estimate the 3D locations of one or more detected plants. It may provide the depth information of selected zapping locations of one or more detected plants determined by the imaging system. Since efficacy for eradicating a weed using high voltage electricity is generally equal in all zapping locations on a targeted plant, as described above in the Zapping Location Determination section, the system is indifferent as to where the zapping contact points are on the targeted plant, the system may randomly select one or more pixels in the region / mask of the plant and use the corresponding depth data to form the zapping location or simply select the pixel with the lowest depth value in the region of the plant mask, which is usually the top of the plant, as the zapping location. In some situations, the system may select the ground depth as the depth value for all zapping locations, as the efficacy for electric weeding on each plant is the same in all locations.

[0160] The depth sensor can be used to estimate the ground depth. It may provide the depth data of a series of lowest points on the 3D map or a group of pixels farthest from the depth sensor in the 3D map to estimate the ground depth. It may also use the imaging system to detect soil texture within the field of view during operations and utilize the corresponding depth data from pixels in the soil region to estimate ground depth.

[0161] In some examples, ground depth information can be acquired through carrier wheels, supporting wheels, probes, physical calibration devices, stereo cameras, LIDAR, radar, time of flight sensor, or sensors that make direct contact with the ground to measure depth accurately. For instance, a wheel measures ground depth by acting as a contact-based reference point that tracks surface variations in real time. As the wheel rolls along the ground, it physically follows terrain contours, allowing the system to determine elevation changes.

[0162] When the growth stage and species of the targeted crop or weed are identified, a plant height value / a height representation is assigned based on historical data from similar plants at corresponding growth stages. This height value / representation is then converted into depth information to assign to critical pixels within the plant masks or bounding boxes after addition of the ground depth measurements.

[0163] In some examples, ground depth information can be acquired through carrier wheels, supporting wheels, probes, physical calibration devices, stereo cameras, LIDAR, radar, time of flight sensor, or sensors that make direct contact with the ground to measure depth accurately. The plant related depth information is not actively collected during operations. When a targeted plant is detected, the control subsystem will set a preprogrammed treatment height value with relation to the ground level and direct the actuator to extend to the treatment height value and conduct treatment. For instance, the treatment height value can be zero, which means the actuator will extend to the ground level. The treatment height value can be 5 centimeters, which means the actuator will extend to ground depth minus 5 centimeters. Since in this case the crop depth data is not collected, the system cannot determine how tall the crops are. If there are crops between the first and second targeted weed, to prevent damaging the crops during operations, after completing a treatment on the first weed, the actuators can be retracted to a pre-determined safety depth before moving on to the second target. The safety depths can be determined through historical data or using the maximum retraction depth for the selected actuators.Execution of Eradication Strategies

[0164] In some examples, the actuators working space overlaps / registers perfectly with the cameras / input system’s field of view. The control system will map the actuators / robot arm position with the input system / camera field of view) which involves establishing a spatial relationship between the actuator’s coordinate system and the camera’s coordinate system, which ensures that objects detected in the camera image correspond to precise locations in the robot's working space. It begins with defining both coordinate systems — the robot arm operates in world coordinates (X, Y, Z, 0), whilethe camera captures data in image coordinates (u, v, depth). A transformation matrix converts image coordinates to real-world coordinates. 3D representations can directly map objects to the actuator’s reference frame. The actuator can determine object positions in the camera image and translate them into real-world coordinates, enabling precise weed eradications, crop manipulation, or tracking. In some examples, to improve synchronizations between the actuators and cameras in high speeds, the actuator controller sends a trigger signal to the camera when the end effector / electrode reaches a predefined position. The camera captures an image exactly at that moment to reduce motion blur and improve accuracy. In some other examples, the camera provides real-time feedback to guide the robot dynamically. The actuators adjust its movement based on camera data, ensuring precise positioning. In some other examples, the camera and actuators operate asynchronously, but each records timestamps. The system matches camera frames to robot positions using timestamps.

[0165] In some examples, the system determines the position of an actuator relative to an object, preferably a plant. One or multiple sensors in the input subsystem may be distributed among any part of the carrier in any combination. The input subsystem can then determine the position of an actuator in relation to an object / plant, and will be configured to identify that object / plant, target it, and / or perform an action or apply a treatment autonomously.IMU (Inertial Measurement Unit) and Uneven Ground and Motions

[0166] In some examples, the input system integrates with an Inertial Measurement Unit (IMU), which tracks motion and orientation using accelerometers, gyroscopes, and sometimes magnetometers. The accelerometer detects linear acceleration and tilt, while the gyroscope measures angular velocity (roll, pitch, and yaw). In some cases, a magnetometer helps correct heading relative to the Earth's magnetic field. IMU data is processed using sensor fusion algorithms like the Kalman filter to provide accurate movement tracking.

[0167] The IMU assists the adaptations to uneven terrain by continuously measuring changes in orientation and acceleration. When a wheel dips into a rut or drives over an obstacle like a large rock, the accelerometer detects the sudden vertical movement or tilt, while the gyroscope captures the change in angular velocity (roll, pitch, and yaw), after combining inputs from multiple sensors to provide a stable and accurate estimate of the carrier system’s orientation. When the input system detects when it is tilting or losing stability, it sends IMU data to adjust robot arms, cameras, and actuators reference points dynamically. To compensate for these irregularities, the control system can use the IMU data and the dimensional and positional data of the carrier system, the actuators, and the sensors to make corrections by adjusting 3D spatial information for pixels in the area of interest to ensure accurate perception and mapping, modifying suspension behavior, and stabilizing onboard actuators. For cameras or depth sensors mounted on the vehicle, IMU data are used to apply real-time stabilization corrections, preventing motion blur or misalignment. For example, if one wheel suddenly drops into a depression, the system can react by quickly adjusting the motions of actuators to maintain even contactwith plants or ground. Similarly, when driving over rough surfaces, the system aligns mounted sensors (such as cameras or depth sensors) using IMU data.Selectively Conductive Electrode

[0168] In some examples, the electrode is configured to be partially conductive and is needle-shaped to minimize the footprint in treating the cluster and overlay situations where the weed grows underneath layers of crops. The electrode can be manufactured with multiple separate shells insulated from each other. The shells assemble to form the complete electrode, with each shell independently connected to the power source through a switch controlled by the control system. One or more electrodes, acting as the end effectors, are physically attached to the end of one actuator. The dimensions of the whole electrode as well as each shell of the electrode is preprogrammed into the control subsystem. After initial calibrations, the spatial positions of the whole electrode as well as each shell of the electrode within the weed eradication system’s working space is known. This system selectively activates specific sections of the electrode surface at precise moments during the treatment process to minimize the false treatment on adjacent plants, such as crops. For instance, when the input subsystem has detected an overlay situation where the weed grew underneath layers of crops, after the control system has identified the situation, it would insert the electrode through the layers of crops and be in contact with the targeted weed with its tip. If the entire surface area of the electrode is conductive, it is appropriate to assume the middle part or lower end of the electrode may be in contact with foliage of crops in the process, which may destroy or damage the crops.

[0169] The advantages of present disclosure is that the electrode can remain non-conductive in the insertion step. Once the tip has reached the weed, only the tip can be switched to conductive and perform the electric treatment, and switched off in the retraction step to avoid false treatment and crop damages.

[0170] The three-dimensional (3D) representation of the areas of interest is crucial to enabling weed eradications in overlay and cluster situations using selectively conductive electrodes. The depth information is used to instruct the control subsystem where the ground is, where the top layers of crops are, and where the weeds are in the depth dimension and in turns instruct the said selectively conductive electrode when and which shell should be turned on and off to successfully eradicate weeds hidden underneath layers of crops, while protecting the top layer crops. For instance, when the said electrode inserts from top to bottom to reach a weed growing close to the ground level and also covered by crops, the input subsystem creates 3D representations for the area of interest with known depth values on critical pixels on the targeted weeds, the covering crops, and the ground. The said electrode will be turned off when the conductive parts (tip, middle, and the lower end) of the electrode reaches the crop height, and turned on the tip when the electrode passes through the crop height and reaches close to the ground level.

[0171] In some examples, one or multiple manipulators in the non-conductive mode may be used to block or push aside the adjacent plants or objects to isolate the targeted plants or objects for othermanipulators to treat. In some examples, the one or multiple manipulators may be used to treat objects on the surface of plants, such as pests.

[0172] FIGS. 5 and 6 are graphical representations of accessed and grid images of a portion of a crop, with weed plants WP and a crop plant CP identified.

[0173] FIG. 7 is a 2D graphical representation 200 of an eradication map showing electrocution points P and an avoidance area AA on a grid. Avoidance area AA may represent an area containing the 2D footprint, or envelope, of a wanted plant such as a crop plant, or of some obstruction that may damage the electrode should it come in contact with it, such as a large rock.

[0174] FIG. 8 is a 3D (three-dimensional) graphical representation of an eradication map 300 showing electrocution points P and an avoidance volume AV. Avoidance volume AV represents a 3D subregion in the overall 3D region proximal to the platform that movement of an electric treatment module as it traverses 3D space according to the eradication map should avoid. Avoidance volume AV may represent a region containing the 3D footprint, or envelope, of a wanted plant such as a crop plant, or of some obstruction that may damage the electrode should it come in contact with it, such as a large rock.3D Eradication Map

[0175] In commercial farms, plants may grow in layers and are generally in various heights. In overlay situations, weeds can grow underneath layers of crops and vice versa. In cluster situations, the plants-of-interest (weed) and plants-of-no-interest (crop) or objects grow close together from the imaging subsystem perspective. Crops and weeds often interleave, occlude, or overlay each other.

[0176] The electric weeding method is less tolerant to crop mistreatment compared to precision chemical weeding and laser weeding. It requires highly precise and selective delivery of electric energy to weeds. To precisely eradicate weeds using electric weeding methods in such complex landscapes in commercial farms, at least one 3D representation of the field is required.

[0177] The disclosed system pertains to the generation of a 3D grid representation, as shown in FIG.8, of the agricultural machinery and its environment, facilitating precise weed and crop mapping within a designated treatment area. Utilizing mapping programs, the system ensures efficient treatment of targeted plants while avoiding crops.

[0178] The output of the plant identification system are high-resolution images, such as represented in FIG. 5 with the objects of interest identified as either weed plants WP or crop plants CP. This system classifies the detected objects into designated map elements, as in FIG. 6. This system then integrates the location information of the weeds and crops with the corresponding spatial depth data to produce an eradication map, as shown in FIG. 7. The utilization of said map allows for the system to generate an optimized path for the treatment process, ensuring efficient and precise movement of the actuators and avoidance of mistreatment on crops or plant-of-no-interest.

[0179] 3D Map generation occurs by taking location information output from the plant identification model, selecting the weed and crop elements. The 2D mapping of plant pixels as in FIG. 6 is combinedwith depth sensor information, including but not limited to LIDAR, RGB-D, stereo depth cameras, TOF (time of flight) cameras, in order to create a 3D representation of the locations of plants including but not limited to point clouds, voxel representations, mesh models, a set of coordinates or other software data structures. This pixel information is then converted to spatial coordinates relative to the location of the agricultural machine’s image sensor. The model would then assign each spatial coordinate its corresponding height in order to construct 3D points representing zap locations and crop boundaries.

[0180] The crop boundaries are determined by the plant identification system, which generates the boundaries taking into account different plant conditions, stages of growth, clustering, and occlusions. Adjustable buffer zones around the crop perimeter can be added to the boundary coordinates to increase crop safety in late stage weeding. This is done by receiving each boundary as a list of pixel points, and generating a new list that represents the new boundary. For each point in the old boundary, the system takes the slope of the two lines connecting to it, and finds the mean slope. Following this trajectory the system can move the padding size distance and generate a corresponding point in the new list. Doing this in order for all points will result in a new list which can be turned into a boundary by connecting adjacent points in the list.

[0181] The resulting collection of coordinates exists in a 3D grid representation of the robot and its workspace. The zapping location coordinates and crop boundaries are then combined into one world representation of the workspace, as in FIG. 8, to guide the route mapping process.

[0182] The eradication map of FIG. 7 serves as a guide for the actuators for delivery of electrical energy, indicating where weed treatment should be applied. In parallel, the eradication map also has crop avoidance information (such as avoidance area AA) indicating locations where the actuators must avoid during the treatment process. This is realized by transforming the crop avoidance information into a series of distances in a three dimensional space that the system interprets in order to move the actuator. The agricultural machinery treats the identified plants according to the generated eradication map with a guarantee to avoid plants.

[0183] With the 3D eradication map in place, the system proceeds to map a path for the actuators to traverse for treatment. The route mapping program considers several critical factors to ensure optimal performance. The route mapping program may optimize the overall travel distance to enhance operational efficiency. In some examples, the route mapping program may strictly adhere to the crop avoidance information to prevent damage or harm to the crops. The route mapping process may also ensure that all weed locations identified in the eradication map are visited and treated effectively.

[0184] In one example, the model computes an optimal path, using obstacle distance information and target treatment points that results in a sequence of movements and actions for the electrical actuators to follow, ensuring comprehensive weed treatment while safeguarding crops. Many factors are taken into account such as speed, distance, acceleration, and external forces. A shortest path for actuators may be calculated, and / or a path that, while not the shortest, may be optimized to reduce overall wearover time of the actuators as compared with other paths that may require several fast stops and starts back and forth in various directions thus imparting wear and tear to the actuation machinery.

[0185] The calculated path is communicated to the electrical actuators, which then execute the treatment tasks. As the actuators operate, real-time feedback from sensors continuously updates the eradication map, ensuring any dynamic changes in the environment are accounted for. This feedback loop enables adaptive path adjustments and enhances the system's robustness and accuracy over a static system by continually and dynamically monitoring for changes and reacting promptly to incoming information.

[0186] FIG. 9 is a 3D graphical representation of an eradication map 300 showing electrocution points P, an avoidance volume AV, and a path 350 on which the electrocution points P sit and that goes around (i.e. avoids) the avoidance volume AV.Electric Applicator Route Mapping

[0187] An electric application route mapping module receives plant location and size information from the eradication map (FIG. 8) as input. In some examples, this input might be a voxel grid, 3D mesh / model, set / range of coordinates, or other software data structures. This information is produced from an imaging subsystem that works with plant data gathered by an input subsystem which may be processed in real time. This electrical applicator route mapping module then determines the path 350 (FIG. 9) that the electrical applicator is to traverse in order to reach all the desired plant identification markers, which may signify one or multiple weeds to eradicate, while simultaneously avoiding touching any crops or other objects to avoid, which is now considered as obstacles in the eradication map represented by avoidance volume(s) (AV). Only one avoidance volume AV is shown in FIGS. 8 and 9 for simplicity of explanation, but a given eradication map may include more than one avoidance volume AV. Also, a given eradication map may, in some instances, include one or more avoidance volume AV but no electrocution points, or may, in other instances, include electrocution points but no avoidance volume AV. That is, whether there is one or more electrocution points P and whether there is one or more avoidance volume AV, will depend on the content of the images captured by the imaging subsystem and how the contents are processed by the processing subsystem to identify and locate the content.

[0188] The control subsystem may achieve control over the actuator(s)’ movement and actuation by initially determining the optimal sequence of weed location markers P to approach. In one example, for each sequential group of targeted plant markers, the control subsystem employs a process to determine the optimal physical path for the electrical applicator to reach the selected targeted plants in the shortest possible time, while avoiding obstacles such as protected plants or physical objects, including stones or small animals. This constitutes a real-time system wherein the motion of the electrical applicator — specifically, the selection of which weed to target and the movement from one weed to the next — is dynamically determined as the entire platform moves and continuously gathersnew plant data. The electrical applicator is, in examples, controlled by various actuators to ensure precise targeting and movement.

[0189] In some examples, the actuator position control and determination is executed through multiple stages. Initially, the group of weed locations and wanted plant locations may be divided into multiple areas with varying priorities which may be based on the motion of the robot and / or the location of the crops to avoid (as some may be in between weeds). The system may determine the required time to move from one location to another, calculated without considering environmental obstacles to reduce computation time. A value may be assigned based on how long a location remains within the robot's reachable workspace, or that of a given electric treatment module in the array, thereby prioritizing locations accessible for shorter durations. Potential collisions with obstacles (both crop and other) may be detected while moving between locations, prioritizing obstacle-free paths to reduce complexity. An obstacle-free path is then established. Based on these metrics, the weed target ordering process may identify the optimal next target and command the robot to proceed to the designated location. While the robot moves towards the assigned location, the weed target ordering process may concurrently determine the subsequent optimal weed to treat. In some examples, a buffer of locations may be maintained in the program memory; as the robot reaches a location, the corresponding point may be removed from the queue. Concurrently, the imaging system may detect new locations and add them to the buffer. That is, in the present system, a first identification of an unwanted plant in an image captured by the imaging subsystem may not correspond to an immediate movement of the actuator to eradicate that unwanted plant. Rather, the first identification of the unwanted plant may serve as the start of a tracking by the processing subsystem of that unwanted plant as it traverses the subsequent images captured by the imaging subsystem and its allocation to a particular set of treatments. For example, if a given unwanted plant can in principle be detected and tracked across 250 frames of digital video given the frame rate and speed of the platform as it traverses a field, it may be put on a particular traversal path of some other actuator in the array, only after several frames containing it have been captured. This may be useful for verifying that the initially -indicated unwanted plant is indeed an unwanted plant, but also useful for ensuring eradication of the unwanted plant can fit into a traversal path of all of the actuators in the array that satisfies various optimization criteria. For example, it may only be after (for example) 100 or 200 or some other number of these frames have been captured and processed that the processing subsystem deems it appropriate to put that tracked unwanted plant on a traversal path for eradication. It will be understood that the processing system may cause the unwanted plant to be put on a traversal path that is not a local shortest path if the end of its treatment window - the window of 3D space in which that tracked unwanted plant can possibly be treated - is approaching and it has not yet been put on a traversal path. As there are multiple applicators, the processing subsystem may determine a different traversal path for each, and switch allocations of a given tracked object from the traversal path of one applicator to the traversal path of another in order to optimize for various otherfactors. Various processes can be deployed to account for object tracking and optimal placement on a traversal path for treatment.

[0190] In some examples, the electrical applicator route mapping process, which is intended to determine the optimal physical path the applicators must take to address one weed and then another while avoiding detected crops or other obstacles, will include a route mapping program, where the imaging subsystem first receives every zapping location in its field of view of and assigns a heuristic value which represents the theoretical cost of reaching that location. This can be a Euclidean straight line distance. In some implementations, the subsystem can calculate the actual cost (duration of travel) of moving the electrical applicator to the neighbouring locations. Based on zapping locations available to move to, the system chooses the one with the smallest cost and heuristic, building a tree of routes until a path is planned for all nodes. Other applications of routes can include but are not limited to RRT (rapidly-exploring random tree), RRT star, and Dijkstra’s algorithm. The 2 or 3-dimensional position of the targeted plant markers and possible obstacle instances will be inputted into this process. The output may be a series of positions that the electrical applicator must follow, otherwise known as a “route”.Weed Eradication Verification

[0191] In examples, the farming machine may be equipped with a verification subsystem that records measurements of the surrounding environment, which are utilized to verily or determine the extent of a previous or ongoing plant treatment. The verification subsystem may measure the geographic area previously assessed by the detection mechanism, encompassing the plants treated by the treatment mechanisms. The verification subsystem may be similar or different from the detection mechanism and can include various sensors such as multispectral cameras, stereo cameras, CCD (charge-coupled device) cameras, single-lens cameras, hyperspectral imaging systems, LIDAR (light detection and ranging) systems, dynamometers, IR (infrared) cameras, thermal cameras, humidity sensors, light sensors, temperature sensors, or any other suitable sensors. In alternative configurations, the verification subsystem can be integrated into other components of the system. The output of this verification subsystem may be a dataset containing information on which sections of the plant area, or field, have been fully treated and which sections may still have weeds present.

[0192] In some examples, the system can also be used to verily plant treatment after an initial electric treatment. For example, a system may capture post-treatment data, which in some cases could be an image. This post-treatment data contains information about objects in the field and any plants treated by the electrical applicator. The verification in the context of electric plant removal involves checking for indicators such as plants releasing smoke, falling down, turning dark, withering, etc., after being treated. The locations of these indicators may be recorded into a data structure (such as a matrix or series of coordinates) and overlaid with saved eradication maps to determine how effective the plant treatment was in a particular session as well as determine if crops are being properly avoided. This verification subsystem may utilize the same data gathering sensors / cameras as the system uses for initialgathering of plant data. This verification subsystem may also be a machine learning model that may be trained to recognize images of post-electrically treated plants.

[0193] In some examples, this verification subsystem may be incorporated into a feedback loop where the targeted plant eradication confirmation data may be transmitted to the controller while plants are being eradicated, for the purpose of informing the controller if certain plants have been missed. The controller can then mark those plants as targeted and may direct the electrical applicator back to them if possible.Carrier Platform

[0194] An electric applicator for imparting electrical treatment can be supported and powered or pulled separately by autonomous vehicles or conventional vehicles operated by a human driver. It can also be implemented on autonomous robots, drones, farming machines / tractors / trailers or other mobile platforms etc.

[0195] The system can additionally function to traverse through a geographic area, such as a field. The system is preferably a ground-based system, but can alternatively be an airborne system or any other suitable system. The system can traverse through the field at a constant, irregular velocity, constant or irregular acceleration, constant plant or feature count, or at any other suitable speed. The system is intended to allow for non-destructive, in-situ plant data collection where the plant stays in place during the process. This data collection and analysis occur in real-time as the system moves through a geographic area.

[0196] The system is configured to detect, identify, and treat plants / objects autonomously using machine learning, deep learning, and other branches of artificial intelligence to assist identification and treatment of objects. The system also incorporates mechanical, electrical and electronic hardware, wired or wireless network communications, and robotics, mobility technologies.

[0197] In some examples, a carrier platform is capable of navigating around the entirety or the majority of a crop, either manually, through pre-trained instructions, semi-autonomously, or fully autonomously, within a designated area. The moving platform may be equipped with maneuvering capabilities facilitated by a combination of wheels and other attachments connected to the platform. The wheels may be governed by the platform's navigation stack. This platform may be configured to be controlled manually as well.

[0198] In some examples, a carrier platform may support a combination of software, hardware, and sensors operating in concert for navigational purposes. The carrier platform may be outfitted with one or more sensors, as part of the input subsystem, to gather sensory information. Sensors may include one or more image capture devices to identify an agricultural object, one or more location or position sensors, one or more odometry sensors, and the like. The position sensors may provide data to determine the locations of an agricultural object relative to a reference coordinate frame or the carrier platform. Sensors can also be configured to detect, analyze, and / or communicate data. Note that any sensor from the input subsystem may be configured to provide sensor data for any purpose. The newlyacquired sensory information, in conjunction with previously collected sensory data, may be merged and fused to yield the requisite information for navigation. This sensory information may be utilized to detect the actual motion of the moving platform within the environment to facilitate autonomous navigation of the carrier platform.

[0199] In some examples, the carrier platform may be configured to generate odometry information. Odometry information would be collected by position and / or location-related sensors which may include wheel encoder or direction sensors, wheel speed sensors, and the like. This odometry information may be employed to adjust the maneuvers of the carrier platform to adhere to the desired path. A navigation subsystem may exploit the stationary nature of visible features within the field of view (FOV) of the imaging system. The spatial relative displacement of objects may remain invariant to the mobile platform's motion. This characteristic may be exploited to estimate the actual movement of the carrier platform based on the relative displacement observed between stationary objects in consecutive data sets collected by the imaging system.

[0200] In some examples, upon completing a task in one region of interest, the carrier platform is moved so as to advance just beyond the current field of view to a new region of interest and stops at the second field of view and forms a new eradication strategy based on the inputs, constructing a new 3D eradication map, mapping new routes for eradication, and directing actuators to conduct the corresponding tasks. This process repeats iteratively until all tasks in the designated field are completed, such as fully weeding an entire farm.

[0201] In some examples, the carrier platform is caused to continue to move without stopping while applying eradication strategies. The imaging subsystem continuously receives the images and depth information from the cameras and depth sensors and performs the detection or segmentation, forming a continuous series of detections as the carrier moves down its path. In order to fuse these detections together, two strategies may be applied. The imaging subsystem and processing subsystem may apply a tracking algorithm that connects the detections across different frames. In this way, once a target is detected, one track identifier (ID) is assigned to the target and tracking the target while it is in the field of view of the cameras and depth sensors. This prevents applying the eradication strategy multiple times on the same target. Then the imaging subsystem and processing subsystem uses the displacement of the target across different frames to predict the instant velocity of the robot and predict the position of the targets instantaneously. Furthermore, since the targets are fixed on the ground and the machine is in motion, the predictions can continuously be applied on the targets even when they are out of the field of view of the imaging subsystem. When the robot is passing through uneven spaces, ex. a rut or over a stone, IMU data is applied to provide corrections to the 3D spatial information for pixels in the area of interest to ensure accurate perception and mapping.

[0202] FIG. 10 is a top perspective view of components of a system 5 for electrocuting unwanted plants amongst a crop, with a cover (not shown) of system 5 removed to show components beneath, according to an example. In this figure, the platform 10, wheels 75, and an array 38 of electric treatment modules(or “applicators”) 40, sixty (60) of which are shown forming multiple rows, and only three (3) of which are called out with respective lead-lines marked as “40... ”. It may be seen that, in this example, array 38 has two main groups, each having thirty (30) applicators, with the two groups arranged generally as mirror images about a middle line of the platform 10. Alternative configurations of arrays of applicators 40 are possible. An imaging subsystem 20 is arranged on platform 10 ahead of the array 38.

[0203] FIG. 11 is a rear view of components of the system 5 for electrocuting unwanted plants amongst a crop as in FIG. 10, according to an example. In this figure, the platform 10, wheels 75, wheel-drive motors 77 associated with respective wheels 75, and the array 38 of electric treatment modules (or “applicators”) 40 are identified.

[0204] FIG. 12 is a side view schematic of the system 5 for electrocuting unwanted plants 100 amongst a crop 99, and physically avoiding wanted plants 102 while doing so, according to an example. In this example, system 5 includes a platform 10 that is positionable amongst crop 99. System 5 also includes an imaging subsystem 20, in this example itself having three (3) imaging devices 26A, 26B, 26C and supported by platform 10 and oriented to capture images of at least a portion of crop 99 that is proximal to platform 10. In this example, a processing subsystem 30 processes images received from imaging subsystem 20 to determine at least locations of unwanted plants (such as weed 100) and wanted plants (such as plant 102) in the images and to transform the locations in the images into actuation locations and avoidance locations in a three-dimensional region proximal to platform 10.

[0205] In this example, there are shown - for ease of understanding - only sixteen (16) representations of z-axis movable electric treatment modules 40 supported by platform 10 and extended to respective actuation or avoidance locations along a z-axis (i.e., in an up-down direction on the page). It can be seen that the first two electric treatment modules 40 (the leftmost modules 40) are at respective positions along the z-direction that correspond to respective actuation locations thereby to contact an unwanted plant 100 (the contacts being shown with asterisks: * that represent electrocutions) while a number of electric treatment modules 40 rearward of the first two electric treatment modules 40 are respectively positioned at different locations along the z-direction to avoidance locations, thereby to avoid physically contacting a wanted plant 102.

[0206] As platform 10 is moved leftward over the crop 99 as in FIG. 12, each individual electric treatment module 40 in array 38, despite being in a fixed x-y position with respect to platform 10, may be individually lifted upwards and away along the z-direction from crop 99 to respective avoidance locations to avoid physical contact with wanted plants, such as plant 102, and may, before or afterwards, be dropped along the z-direction to actuation locations so as to make physical contact with unwanted plants, such as plant 100, in order to impart an electrocution to such unwanted plants. It can be seen in FIG. 12 that a number of the electric treatment modules 40 are controlled to collectively form a kind of negative three-dimensional space 105 - an envelope - around wanted plant 102 as the platform advances leftward (only two dimensions are shown in FIG. 12, but a third dimension corresponding to additional electric treatment modules 40 in array 38 would be into the page). The individual lifting and loweringof individual electric treatment modules 40 is conducted by a control subsystem 50, which is operable to move each of the electric treatment modules 40 in the array 38 along the z-direction between the respective avoidance and actuation locations. This is done by control subsystem 50 in accordance with a processing subsystem 30 with which it is in communication that processes images received from an imaging subsystem 20 that is oriented to capture images of at least a portion of the crop 99 that is proximal to platform 10.

[0207] More particular, the processing system 30 processes the images received from the imaging subsystem 20 to determine at least locations of wanted and unwanted plants in the images, and to transform the locations in the images, respectively, into avoidance and actuation locations in the three-dimensional region proximal to platform 10. Generally -speaking, the electric treatment modules 40 are collectively controlled to move in the z-direction to form the negative space 105 to physically avoid each wanted plants 102 as they are approached, and are controlled to move in the z-direction to eliminate again such negative spaces 105 in order to be in a position to physically contact unwanted plants 100 as they are approached. It will be appreciated that, in a given treatment of a crop, electric treatment modules 40 will each be controlled to raise and lower along the z-direction with respect to the platform, likely many times each, over the course of the treatment in order to physically avoid wanted plants and to contact unwanted plants for electrocution.

[0208] In this example, system 5 is autonomous and includes a drive subsystem 80 that drives and steers wheels 75 using individual wheel-drive motors 77 for moving platform 10 amongst and through crop 99. The control subsystem 30 moves the at least one electric treatment module 40 to the actuation locations in coordination with moving of platform 10 by drive subsystem 80.

[0209] In this example, the imaging subsystem 20 includes an artificial lighting subsystem 28 having two lighting units directed within the field of view of the imaging subsystem 20, for imparting artificial light to at least the portion of crop 99 imaged by imaging subsystem 20.

[0210] In this example, each imaging device 26A, 26B, 26C is a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0211] In this example, a depth-sensing subsystem 25 is supported by platform 10 and is oriented to provide depth measurements for use by processing subsystem 30 to transform the locations in the images captured by imaging subsystem 20 into actuation locations or avoidance locations in the three-dimensional region proximal to platform 10.

[0212] System 5 can be operated in various ways, appropriate to the conditions of the field, the crop, efficiency considerations, accuracy considerations, and / or cost considerations, and the like. For example, processing subsystem 30 may processes the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem 50 is operable to determine a format of the time-limited electrocution treatment based on the at least one additional attribute.

[0213] One or more of the at least one additional attribute may be an attribute of an unwanted plant, for example an attribute selected from the group consisting of: moisture level of an unwanted plant, sizeof an unwanted plant, age of an unwanted plant, species of an unwanted plant, or some other attribute of the unwanted plant. One or more of the at least one additional attribute may be an attribute of the crop, for example an attribute selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop, or some other attribute of the crop.

[0214] System 5 may be operated such that control subsystem 50 is operable to determine a format of the time-limited electrocution treatment by selecting the format from a set of formats based on the at least one additional attribute. The formats may be stored as in computer-readable memory as respective sets of parameters. Alternatively or in some combination, control subsystem 50 is operable to determine a format of the time-limited electrocution treatment by creating the format based on the at least one additional attribute. The format may be created by selecting different parameters for combination, or by calculating parameters based on the at least one additional attribute.

[0215] System 5 may be operated such that the format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant, or some other parameter(s).

[0216] System 5 may be operated such that processing subsystem 30 processes the images received from imaging subsystem 20 to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart additional avoidance location (such as an avoidance volume AV), in the three-dimensional region proximal to platform 10. This may be done in order to avoid objects other than plants, such as rocks or piles of soil, so as to avoid damage to electric treatment modules 40 that might otherwise occur were they to contact the rocks or soil. Control subsystem 50 may accordingly be operated to move the each electric treatment module 40 in the array 38 to one or more of the actuation locations in the three-dimensional region while avoiding physically coinciding with any such additional avoidance location.

[0217] In this example, processing subsystem 30 includes a machine-learning model that has been trained, and that may continue to be trained through continued detection and operation from use and / or updates from external sources, to segment unwanted plants in the images and wanted plants in the images thereby to determine at least the locations of the unwanted plants and the locations of the wanted plants.

[0218] System 5 may be operated such that the processing subsystem 5 processes the images received from the imaging subsystem to determine at least locations in the images at which the time -limited electrocution treatment has already been delivered. This may be useful for preparing reports on the efficacy of eradication (or other treatment as described herein) or for ensuring that treatment is not applied repeatedly when not required.

[0219] System 5 may be operated to transform the locations in the images into actuation locations and avoidance locations in the three-dimensional region proximal to platform 10 by generating at least one three-dimensional eradication map (such as eradication map 300) comprising the actuation locations(points P) and avoidance locations in the three-dimensional region for a set of time-limited electrocution treatments. Control subsystem 50 is then caused to move each electric treatment module 40 to respective actuation locations P or avoidance locations based on the three-dimensional eradication map 300.

[0220] It will be appreciated that control subsystem 50 may employ logic to move each electric treatment module 40 along the z-axis only just as far out of the way of a wanted plant as it needs to be to avoid physical contact with the wanted plant. That is, instead of moving a given electric treatment module 40 to its highest z-position when it is required to avoid a wanted plant, and thus possibly quite far away from the wanted plant (especially while the wanted plant is short in height), it may be necessary only to move the electric treatment module to a position that is just above and not touching the wanted plant. By withdrawing the electric treatment module only just as far as it needs to be withdrawn and no farther (except, perhaps, for a given tolerance amount), the electric treatment module may be more likely to be more readily brought into contact with an unwanted plant that follows the wanted plant without having to extend back down all the way from its highest z-position. Such logic may enable the overall system 5 to reduce the total distance each electric treatment module travels back and forth during a crop treatment session, which may reduce wear and tear, but may also reduce the time required for each electric treatment module to reach its directed actuation location after it has been in a (nearby) avoidance location, due to it only having to travel a short distance. By reducing the time required for each electric treatment module to reach its directed actuation location after having been in an avoidance location, it may be possible to operate system 5 so as to move quickly over a crop while also timely conducting the necessary electrocutions of unwanted plants as it does so.

[0221] There may generally be several eradication maps generated based on the differences over time in the input data, such that each three-dimensional eradication map corresponds to a respective treatment time window.

[0222] System 5 may be operated such that a given actuation location is coincident with the unwanted plants. System 5 may be operated such that a given actuation location is proximal to the unwanted plants such that they are near to but do not physically touch the unwanted plants. This may be possible, for example, if the electrical power conditions are such that a circuit can be closed thus passing current through the unwanted plant without the electrode itself physically touching the unwanted plant. This may be useful if a given actuation location, and a given treatment, can deliver electrocution to two or more unwanted plants that are near to each other but that cannot both be physically contacted at the same time by an electrode. That is, it may be useful to increase the power delivery to a particular electrode during delivery of a treatment above that required for eradication one unwanted plant, in order to both eradicate that unwanted plant and eradicate another unwanted plant that is proximal to the electrode, with a single treatment. Variations are possible.

[0223] FIG. 13 is a flow diagram showing various stages in a process 500 of detecting and eradicating weeds using system 5, according to one example. In box 510, an input subsystem collects plant dataand images passively while the overall system 5 is in motion. In box 570, based on input subsystem data, a fitting eradication strategy is chosen or constructed. In box 520, an image captured and containing a plant in a field is accessed and, in box 530, a deep learning model segments the identified plants as either crop instances or weed instances. In box 540, continuous weed (unwanted plant) tracking is conducted to keep track of each instance of weeds, and continuous wanted plant tracking is conducted to keep track of each instance of wanted plant. In box 550 an eradication map is created based on segmented plant instances and, at box 560 optimal platform / actuator movement to each weed is determined using a method such as route mapping. The chosen eradication strategy of box 570 is provided as input to the determinations being made in box 560. The chosen eradication strategy of box 570 is also provided as input to the application of a zapping location algorithm of box 580 that chooses the zapping points in each segment. In box 590 the weed eradication process is executed to conduct approaching along the z-axis, eradicating through electrocution, and moving to a next unwanted plant (weed) possibly after moving first to one or more avoidance locations. In box 595, weed eradication is verified. Alternative processes for each of the boxes shown in FIG. 13, including combinations of multiple of such processes into a single process, are possible.

[0224] FIG. 14 is a bottom perspective view of an electric treatment module 40, according to an example, supported by an actuator 400. FIG. 15 is a side elevation view of the electric treatment module 40, and FIG. 16 is a top view of a transmission system for the electric treatment module 40.

[0225] Actuator 400 includes a frame 410 that supports a motor 420, in this example a servo motor, and a transmission 430 for converting rotational motion of the motor 420 into linear motion, in the z-direction, of the electric treatment module 40 with respect to frame 410. In an example, a servo drive (not shown) is be connected in series between the servo motor and the control subsystem in order to turn low voltage power signals from the control system into high power voltage and currents for operating the servo motor accordingly. The motor 420 is secured to frame 410 and frame 410, in turn, is secured either directly or indirectly to the platform, in a manner that reduces or eliminates forces and vibrations in all axes from being imparted to the actuator 400 and electric treatment module 40.

[0226] In examples, the servo drive includes a brake capacitor to receive and store some of the energy generated by braking of the tine system.

[0227] The servo motor can track and assign rotational positions for accuracy and reliability, translated by the transmission 430 into accurate and reliable linear positions in the z-direction of electrical treatment module 40 and, in particular, its electrode 44.

[0228] In other examples, different kinds of motors, such as solenoids and / or linear motors, may be used, whether with or without a transmission. In some examples, different kinds of transmissions may be used to convert the motion of alternative motors into other linear motions, torsional motions, spiral motions, rotational motions, and other motions.

[0229] In some examples, where the transmission is concerned, ball screws may be used to translate rotational motion into linear motion of the electric treatment modules with high efficiency andprecision. Other methods of providing linear motion are possible, such as linear motors and linear actuators. Each actuator may have one or more sensor for sensing the position of a ball screw. In this example, each actuator has multiple sensors 440 for monitoring the position of the ball screw by detecting the position of a rib 450 that rises or lowers within a slot 455 according to the z-position of the electric treatment module 40 with respect to the actuator 400. These sensors 440 may be used for setting soft or hard stops so that the linear motion can be stopped before hitting the end of its desired range of motion, thereby to prevent damage to the actuator and / or the electric treatment module it is supporting and causing to be moved.

[0230] Rib 450 sliding within slot 455 also prevents rotation of the ball nut, and a fixed support (not shown) prevents both axial and radial forces from moving the ball screw unintentionally.

[0231] In examples, for mechanical advantage as between the motor 420 and the electric treatment module 40, as part of the transmission there may be a pulley system. For example, a timing pulley 460 and belt 470 (see FIG. 16) which allow for mechanical advantage. That is, it is useful to obtain more rotations per minute (rpm) than the servo motor 420 will typically allow some accommodation as to torque can be taken advantage of. Shaft couplings can be used to mount pulley (s). Other methods of changing the direction of forces or multiplying the forces applied are possible, such as gear systems.

[0232] One or more sensors may be used for homing the position of the tine. To home the tine, the system will lower until either reaching the homing sensor or a soft stop sensor. If the soft stop sensor is reached, the tine will raise until reaching the homing sensor. The homing sensor will be at a known distance from the motor and this information may be used to accurately calibrate the position of the tine.

[0233] In examples, a feedback system is made up of a spring-damper system as well as an encoder. The encoder may measure a distance between the tine’s current position and the position where the spring is the most compressed. The encoder may be used to determine when the electrode has made contact with the ground or a rock since the distance between the electrode and the end-effector will have decreased. At this point the control subsystem will actuate the actuator to cause the tine to rise in the z-direction, thereby to avoid or at least reduce damage.

[0234] In examples, the electrode 44 at the end of the tine is insulated from most directions except for the area at which the tine is intended to make contact with the weed or other unwanted plant. This way, it can be ensured that the electrical discharge is solely concentrated on the unwanted plants and will reduce or prevent the instances of energy dissipation into surrounding areas, which could unintentionally damage wanted plants or cause other problems. Electrode 44 is made of a conductive material on the inside which can provide the electrical current to the weed it is trying to electrocute as well as an insulator on the outside so that the current and voltage can be controlled and directed as intended rather than dissipating by hitting wanted plants or other non-targeted plants.

[0235] In examples, an overcurrent protection measure reduces or prevents electrical overload. Such an overcurrent protection measure may be a circuit breaker or a fuse.

[0236] While embodiments described herein are primarily directed to electric treatment of unwanted plants by electrocution, the principles of image capture, processing, control over movement, and actuation may be used in the context of mechanical weeding, where an end effector or other component of an applicator harms the unwanted plant by physical means rather than by electrical means.

[0237] FIG. 17 is a flowchart depicting a method 600 for eradicating weeds amongst a crop, according to an example. Method 600 includes positioning a platform amongst the crop (step 610), capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform (step 620), processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform (step 630); individually moving electric treatment modules that are supported by the platform in a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant (step 640); and individually actuating each of the electric treatment modules while at respective actuation locations thereby to deliver an eradication treatment to the unwanted plants (step 650).

[0238] In an example, the method further comprises processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the eradication treatment is based on the at least one additional attribute.

[0239] In an example, the method further comprises determining a format of the eradication treatment by selecting the format from a set of formats based on the at least one additional attribute.

[0240] In an example, the method further comprises determining a format of the eradication treatment by creating the format based on the at least one additional attribute.

[0241] In an example, the at least one additional attribute is an attribute of an unwanted plant.

[0242] In an example, the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0243] In an example, the at least one additional attribute is an attribute of the crop.

[0244] In an example, the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0245] In an example, the eradication treatment is a time-limited electrocution treatment and a format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant.

[0246] In an example, the method further comprises processing the images to determine at least one additional location of non-contact in the images; transforming each additional location of non-contactin the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; and moving the electric treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

[0247] In an example, the method further comprises moving the platform amongst and through the crop; and moving the electric treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

[0248] In an example, the imaging subsystem comprises at least one image capture device configured to capture respective images.

[0249] In an example, the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0250] In an example, the method further comprises imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0251] In an example, the method further comprises training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

[0252] In an example, the method further comprises processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

[0253] In an example, the method further comprises transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of eradication treatments, wherein the electric treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0254] In an example, the method further comprises providing, by a depth-sensing subsystem supported by the platform, depth measurements; and using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0255] In an example, the actuation locations are coincident with the unwanted plants.

[0256] In an example, the actuation locations are proximal to the unwanted plants.

[0257] While embodiments described herein are primarily directed to electric treatment of unwanted plants by electrocution, the principles of image capture, processing, control over movement, and actuation may be used in the context of mechanical weeding, where an end effector or other component of an applicator harms the unwanted plant by physical means rather than by electrical means.

[0258] For example, a system for mechanically harming unwanted plants amongst a crop may be provided. Such a system may include: a platform positionable amongst the crop; an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximalto the platform; a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; a mechanical treatment subsystem supported by the platform, the mechanical treatment subsystem comprising an array of mechanical treatment modules extending across the platform, each of the mechanical treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually actuated to execute a mechanical harming action; and a control subsystem operable to move each of the mechanical treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually actuate each of the mechanical treatment modules while at the respective actuation location thereby to execute the mechanical harming action to the unwanted plants. Such mechanical harming / treatment may be one or more of: grasping a portion of the targeted plants with an end effector and pulling them out, cutting their stems with a blade of an end effector, scooping them out of soil with a scoop of an end effector, or a combination of these mechanical treatment approaches.

[0259] In an example, the processing subsystem may process the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem is operable to determine a format of the mechanical harming action based on the at least one additional attribute.

[0260] In an example, the control subsystem may be operable to determine a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

[0261] In an example, the control subsystem may be operable to determine a format of the mechanical harming action by creating the format based on the at least one additional attribute.

[0262] In an example, the at least one additional attribute may be an attribute of an unwanted plant.

[0263] In an example, the at least one additional attribute may be selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0264] In an example, the at least one additional attribute may be an attribute of the crop.

[0265] In an example, the at least one additional attribute may be selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0266] In an example, the format of the mechanical harming action may comprise one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

[0267] In an example, the processing subsystem may process the images received from the imaging subsystem to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart avoidance location in the three-dimensional region proximal to the platform. The control subsystem may be operable to move the at least one mechanical treatment module to one or more of the actuation locations in the three-dimensional region while avoiding coinciding with any avoidance location.

[0268] In an example, the system may further comprise a drive subsystem for moving the platform amongst and through the crop, wherein the control subsystem may move the at least one mechanical treatment module to the actuation locations in coordination with moving of the platform by the drive subsystem.

[0269] In an example, the imaging subsystem may comprise at least one image capture device configured to capture respective images.

[0270] In an example, the at least one image capture device may be a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0271] In an example, the imaging subsystem may comprise an artificial lighting subsystem for imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0272] In an example, the processing subsystem may comprise a machine-learning model trained to segment unwanted plants in the images thereby to determine at least the locations of the unwanted plants.

[0273] In an example, the processing subsystem may be configured to process the images received from the imaging subsystem to determine at least locations in the images at which the mechanical harming action has already been delivered.

[0274] In an example, the processing subsystem may transform the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions. The control subsystem may move the at least one mechanical treatment module to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0275] In an example, the system may comprise a depth-sensing subsystem supported by the platform and oriented to provide depth measurements for use by the processing subsystem to transform the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0276] In an example, the actuation locations may be coincident with the unwanted plants.

[0277] In an example, the actuation locations may be proximal to the unwanted plants. For example, a given end effector may have a region of actuation that extends laterally outwards or otherwise away from the actuation location. In particular, a given end effector may include a blade extending from a central axis that can be swung laterally with respect to the central axis to cut an unwanted plant that isnot on the central axis itself. Or, a weed-trimmer style wire may extend a distance from such a central axis and be spun about the central axis during actuation to itself contact the unwanted plant and cut it, such that the actuation location itself (the central axis location, for example) is not coincident with the unwanted plant, but the wire itself can be swung to be coincident with the unwanted plant. A chemical treatment may be applied in a similar manner, where a valve is oriented to spray a liquid form of chemical laterally a short distance with respect to a central axis, such that the central axis corresponds to the actuation location is spaced from the plant itself due to the sprayed liquid chemical being able to traverse the space.

[0278] In accordance with mechanical treatment instead of electric treatment, in an example there may be provided a method for eradicating weeds amongst a crop, the method comprising: positioning a platform amongst the crop; capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform; processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform; individually moving mechanical treatment modules that are supported by the platform in at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; and individually actuating each of the mechanical treatment modules while at respective actuation locations thereby to deliver a mechanical harming action to the unwanted plants.

[0279] In examples, the method may further comprise processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the mechanical harming action may be based on the at least one additional attribute.

[0280] In examples, the method may further comprise determining a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

[0281] In examples, the method may further comprise determining a format of the mechanical harming action by creating the format based on the at least one additional attribute.

[0282] In examples, the at least one additional attribute may be an attribute of an unwanted plant.

[0283] In examples, the at least one additional attribute may be selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0284] In examples, the at least one additional attribute may be an attribute of the crop.

[0285] In examples, the at least one additional attribute may be selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0286] In examples, format of the mechanical harming action may comprise one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

[0287] In examples, the method may further comprise processing the images to determine at least one additional location of non-contact in the images; transforming each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; and moving the mechanical treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

[0288] In examples, the method may further comprise moving the platform amongst and through the crop; and moving the mechanical treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

[0289] In examples, the imaging subsystem may comprise at least one image capture device configmed to capture respective images.

[0290] In examples, the at least one image capture device may be a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0291] In examples, the method may further comprise imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0292] In examples, the method may further comprise training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

[0293] In examples, the method may further comprise processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

[0294] In examples, the method may further comprise transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions, wherein the mechanical treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0295] In examples, the method may further comprise providing, by a depth-sensing subsystem supported by the platform, depth measurements; and using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0296] In examples, the actuation locations may be coincident with the unwanted plants.

[0297] In examples, the actuation locations may be proximal to the unwanted plants.

[0298] FIG. 18 is a schematic diagram showing a hardware architecture of a computing system 1000. Computing system 1000 is suitable as the hardware platform for components of system 5, such as processing subsystem 30 and / or components of control subsystem 50, drive subsystem 80, and / or other components.

[0299] Computing system 1000 includes a bus 1010 or other communication mechanism for communicating information, and a processor 1018 coupled with the bus 1010 for processing the information. The computing system 1000 also includes a main memory 1004, such as a random access memory (RAM) or other dynamic storage device (e.g., dynamic RAM (DRAM), static RAM (SRAM), and synchronous DRAM (SDRAM)), coupled to the bus 1010 for storing information and instructions to be executed by processor 1018. In addition, the main memory 1004 may be used for storing temporary variables or other intermediate information during the execution of instructions by the processor 1018. Processor 1018 may include memory structures such as registers for storing such temporary variables or other intermediate information during execution of instructions. The computing system 1000 further includes a read only memory (ROM) 1006 or other static storage device (e.g., programmable ROM (PROM), erasable PROM (EPROM), and electrically erasable PROM (EEPROM)) coupled to the bus 1010 for storing static information and instructions for the processor 1018.

[0300] Computing system 1000 also includes a disk controller 1008 coupled to the bus 1010 to control one or more storage devices for storing information and instructions, such as a magnetic hard disk 1022 and / or a solid state drive (SSD) and / or a flash drive, and a removable media drive 1024 (e.g., solid state drive such as USB (universal serial bus) key or external hard drive, floppy disk drive, read-only compact disc drive, read / write compact disc drive, compact disc jukebox, tape drive, and removable magnetooptical drive). The storage devices may be added to the computing system 1000 using an appropriate device interface (e.g., Serial ATA (SATA), peripheral component interconnect (PCI), small computing system interface (SCSI), integrated device electronics (IDE), enhanced-IDE (E-IDE), direct memory access (DMA), ultra-DMA, as well as cloud-based device interfaces).

[0301] Computing system 1000 may also include special purpose logic devices (e.g., application specific integrated circuits (ASICs)) or configurable logic devices (e.g., simple programmable logic devices (SPLDs), complex programmable logic devices (CPLDs), and field programmable gate arrays (FPGAs)).

[0302] Computing system 1000 also includes a display controller 1002 coupled to the bus 1010 to control a display 1012, such as an LED (light emitting diode) screen, organic LED (OLED) screen, liquid crystal display (LCD) screen or some other device suitable for displaying information to a computer user. In embodiments, display controller 1002 incorporates a dedicated graphics-processing unit (GPU) for processing mainly graphics-intensive or other parallel operations. Such operations may include rendering by applying texturing, shading and the like to wireframe objects including polygons such as spheres and cubes thereby to relieve processor 1018 of having to undertake such intensiveoperations at the expense of overall performance of computing system 1000. The GPU may incorporate dedicated graphics memory for storing data generated during its operations, and includes a frame buffer RAM memory for storing processing results as bitmaps to be used to activate pixels of display 1012. The GPU may be instructed to undertake various operations by applications running on computing system 1000 using a graphics-directed application-programming interface (API) such as OpenGL, Directs D and the like.

[0303] Computing system 1000 includes input devices, such as a keyboard 1014 and a pointing device 1016, for interacting with a computer user and providing information to the processor 1018. The pointing device 1016, for example, may be a mouse, a trackball, or a pointing stick for communicating direction information and command selections to the processor 1018 and for controlling cursor movement on the display 1012. The computing system 1000 may employ a display device that is coupled with an input device, such as a touch screen. Other input devices may be employed, such as those that provide data to the computing system via wires or wirelessly, such as gesture detectors including infrared detectors, gyroscopes, accelerometers, other kinds of input devices such as radar / sonar, front and / or rear cameras, infrared sensors, ultrasonic sensors, LIDAR (Light Detection and Ranging) sensors, and other kinds of sensors.

[0304] Computing system 1000 performs a portion or all of the processing steps discussed herein in response to the processor 1018 and / or GPU of display controller 1002 executing one or more sequences of one or more instructions contained in a memory, such as the main memory 1004. Such instructions may be read into the main memory 1004 from another processor readable medium, such as a hard disk 1022 or a removable media drive 1024. One or more processors in a multi-processing arrangement such as computing system 1000 having both a central processing unit and one or more graphics processing unit may also be employed to execute the sequences of instructions contained in main memory 1004 or in dedicated graphics memory of the GPU. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions.

[0305] As stated above, computing system 1000 includes at least one processor readable medium or memory for holding instructions programmed according to the teachings of the description and for containing data structures, tables, records, or other data described herein. Examples of processor readable media are solid state devices (SSD), flash-based drives, compact discs, hard disks, floppy disks, tape, magneto-optical disks, PROMs (EPROM, EEPROM, flash EPROM), DRAM, SRAM, SDRAM, or any other magnetic medium, compact discs (e.g., CD-ROM), or any other optical medium, punch cards, paper tape, or other physical medium with patterns of holes, a carrier wave (described below), or any other medium from which a computer can read.

[0306] Stored on any one or on a combination of processor readable media, is software for controlling the computing system 1000, for driving a device or devices to perform the functions discussed herein, and for enabling computing system 1000 to interact with a human user. Such software may include, but is not limited to, device drivers, operating systems, development tools, and applications software.Such processor readable media further includes the computer program product for performing all or a portion (if processing is distributed) of the processing performed discussed herein.

[0307] The computer code devices discussed herein may be any interpretable or executable code mechanism, including but not limited to scripts, interpretable programs, dynamic link libraries (DLLs), object-oriented programming (OOP) modules such as classes, and complete executable programs. Moreover, parts of the processing of the present description may be distributed for better performance, reliability, and / or cost.

[0308] A processor readable medium providing instructions to a processor 1018 may take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. Non-volatile media includes, for example, optical, magnetic disks, and magneto-optical disks, such as the hard disk 1022 or the removable media drive 1024. Volatile media includes dynamic memory, such as the main memory 1004. Transmission media includes coaxial cables, copper wire and fiber optics, including the wires that make up the bus 1010. Transmission media also may also take the form of acoustic or light waves, such as those generated during radio wave and infrared data communications using various communications protocols.

[0309] Various forms of processor readable media may be involved in carrying out one or more sequences of one or more instructions to processor 1018 for execution. For example, the instructions may initially be carried on a magnetic disk of a remote computer. The remote computer can load the instructions for implementing all or a portion of the present description remotely into a dynamic memory and send the instructions over a wired or wireless connection using a modem. A modem local to the computing system 1000 may receive the data via wired Ethernet or wirelessly via Wi-Fi and place the data on the bus 1010. The bus 1010 carries the data to the main memory 1004, from which the processor 1018 retrieves and executes the instructions. The instructions received by the main memory 1004 may optionally be stored on storage device 1022 or 1024 either before or after execution by processor 1018.

[0310] Computing system 1000 also includes a communication interface 1020 coupled to the bus 1010. The communication interface 1020 provides a two-way data communication coupling to a network link that is connected to, for example, a local area network (LAN) 1500, or to another communications network 2000 such as the Internet. For example, the communication interface 1020 may be a network interface card to attach to any packet switched LAN. As another example, the communication interface 1020 may be an asymmetric digital subscriber line (ADSL) card, an integrated services digital network (ISDN) card or a modem to provide a data communication connection to a corresponding type of communications line. Wireless links may also be implemented. In any such implementation, the communication interface 1020 sends and receives electrical, electromagnetic or optical signals that carry digital data streams representing various types of information.

[0311] The network link typically provides data communication through one or more networks to other data devices, including without limitation to enable the flow of electronic information. For example,the network link may provide a connection to another computer through a local network 1500 (e.g., a LAN) or through equipment operated by a service provider, which provides communication services through a communications network 2000. The local network 1500 and the communications network 2000 use, for example, electrical, electromagnetic, or optical signals that carry digital data streams, and the associated physical layer (e.g., CAT 5 cable, coaxial cable, optical fiber, etc.). The signals through the various networks and the signals on the network link and through the communication interface 1020, which carry the digital data to and from the computing system 1000, may be implemented in baseband signals, or carrier wave based signals. The baseband signals convey the digital data as unmodulated electrical pulses that are descriptive of a stream of digital data bits, where the term "bits" is to be construed broadly to mean symbol, where each symbol conveys at least one or more information bits. The digital data may also be used to modulate a carrier wave, such as with amplitude, phase and / or frequency shift keyed signals that are propagated over a conductive media, or transmitted as electromagnetic waves through a propagation medium. Thus, the digital data may be sent as unmodulated baseband data through a "wired" communication channel and / or sent within a predetermined frequency band, different from baseband, by modulating a carrier wave. The computing system 1000 can transmit and receive data, including program code, through the network(s) 1500 and 2000, the network link and the communication interface 1020. Moreover, the network link may provide a connection through a LAN 1500 to a mobile device 1300 such as a personal digital assistant (PDA) laptop computer, or cellular telephone.

[0312] Alternative configurations of computing systems may be used to implement the systems and processes described herein.

[0313] Electronic data stores implemented in the database described herein may be one or more of a table, an array, a database, a structured data file, an XML file, or some other functional data store, such as hard disk 1022 or removable media 1024.

[0314] Although examples have been described, those of skill in the art will appreciate that variations and modifications may be made without departing from the spirit, scope and purpose of the invention as defined by the appended claims.

[0315] For example, while in examples described herein, the treatment modules (for example, the electrical treatment modules, the mechanical chemical treatment modules, or the chemical treatment modules) of the arrays are each fixed in the x-y position with respect to the platform, and thus movable in only the z-direction, alternatives are possible. For example, in alternative examples each, or some, given treatment modules in the array may be movable in the x direction and / or the y direction so as to be moved slightly in these directions but within a respective constrained region. As described herein, Tao 1 describes a system in which one treatment module, or a very small number of treatment modules such as two treatment modules, are each caused to be moved extensively in all of the x-y-z directions as treatment progresses. However, a compromise between array configurations that permit only a fixed x-y position for each treatment module, and the Tao 1 configuration, offering high-resolution treatmentwith less wear and tear and possibly more time-efficient operation, may be provided by enabling each treatment module in a tightly packed array to itself be moved in the x-y direction within a respective small subregion. For example, a given treatment module may be eccentrically shaped about its z-axis and may be rotatable by a motor system about that z-axis such that its tip - such as its electrode - can reach to different locations within its small subregion. More particularly, if a given tip of an electrode is only 1 centimeter in diameter, it may nevertheless be configured to be rotated back and forth about the z-axis so that the tip can cover portions of a larger subregion - a larger-diameter cylinder-shaped subregion, for example - than just the 1 centimeter tip would “cover”. Other such simple mechanisms for enabling the treatment module to traverse its own small subregion are possible, such as mechanisms that move the treatment module between two fixed x positions in its subregion so that the control subsystem can shift the treatment module leftward or rightward to an avoidance location or an actuation location as the platform advances in the y direction. Generally -speaking, providing x and / or y movability within a respective small subregion (in addition to the z movability described herein), may enable the resolution or effective density of coverage of treatment modules to be increased without necessarily increasing the discrete numbers of individual treatment modules.

[0316] In examples, a respective small subregion does not overlap any other small subregion of another treatment module; that is the respective subregions are non-overlapping. In examples, at least one of the treatment modules in the array is movable in the x and / or y directions in its respective subregion. In examples, each of the treatment modules in the array is moveable in the x and / or y directions in its respective subregion.

[0317] Clauses. The following sets out various examples in the form of clauses.

[0318] Clause 1. A system for electrocuting unwanted plants amongst a crop, the system comprising:

[0319] a platform positionable amongst the crop;

[0320] an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform;

[0321] a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;

[0322] an electric treatment subsystem supported by the platform, the electric treatment subsystem comprising an array of electric treatment modules extending across the platform, each of the electric treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually electrically actuated to deliver a time-limited electrocution treatment; and

[0323] a control subsystem operable to move each of the electric treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and,while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually electrically actuate each of the electric treatment modules while at the respective actuation location thereby to deliver the time-limited electrocution treatment to the unwanted plants.

[0324] Clause 2. The system of clause 1, wherein the processing subsystem processes the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment based on the at least one additional attribute.

[0325] Clause 3. The system of clause 2, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment by selecting the format from a set of formats based on the at least one additional attribute.

[0326] Clause 4. The system of clause 2, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment by creating the format based on the at least one additional attribute.

[0327] Clause 5. The system of clause 2, wherein the at least one additional attribute is an attribute of an unwanted plant.

[0328] Clause 6. The system of clause 5, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0329] Clause 7. The system of clause 2, wherein the at least one additional attribute is an attribute of the crop.

[0330] Clause 8. The system of clause 7, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0331] Clause 9. The system of clause 2, wherein the format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant.

[0332] Clause 10. The system of any one of clauses 1-9, wherein:

[0333] the processing subsystem processes the images received from the imaging subsystem to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform, and

[0334] wherein the control subsystem is operable to move the electric treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

[0335] Clause 11. The system of any one of clauses 1-10, further comprising:

[0336] a drive subsystem for moving the platform amongst and through the crop, wherein the control subsystem moves the electric treatment modules between the respective avoidance and actuation locations in coordination with moving of the platform by the drive subsystem.

[0337] Clause 12. The system of any one of clauses 1-11, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

[0338] Clause 13. The system of clause 12, wherein the at least one image capture device is a video capture device, and the respective images are frames of a digital video captured by the video capture device.

[0339] Clause 14. The system of any one of clauses 1-13, wherein the imaging subsystem comprises an artificial lighting subsystem for imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0340] Clause 15. The system of any one of clauses 1-14, wherein the processing subsystem comprises a machine-learning model trained to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

[0341] Clause 16. The system of any one of clauses 1-15, wherein the processing subsystem is configured to process the images received from the imaging subsystem to determine at least locations in the images at which the time-limited electrocution treatment has already been delivered.

[0342] Clause 17. The system of any one of clauses 1-16, wherein the processing subsystem transforms the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of time-limited electrocution treatments, wherein the control subsystem moves each of the electric treatment modules to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0343] Clause 18. The system of any one of clauses 1-17, further comprising:

[0344] a depth-sensing subsystem supported by the platform and oriented to provide depth measurements for use by the processing subsystem to transform the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0345] Clause 19. The system of any one of clauses 1-18, wherein the actuation locations are coincident with the unwanted plants.

[0346] Clause 20. The system of any one of clauses 1-18, wherein the actuation locations are proximal to the unwanted plants.

[0347] Clause 21. A method for eradicating weeds amongst a crop, the method comprising:

[0348] positioning a platform amongst the crop;

[0349] capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform;

[0350] processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;

[0351] individually moving electric treatment modules that are supported by the platform in at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; and

[0352] individually actuating each of the electric treatment modules while at respective actuation locations thereby to deliver an eradication treatment to the unwanted plants.

[0353] Clause 22. The method of clause 21, further comprising:

[0354] processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the eradication treatment is based on the at least one additional attribute.

[0355] Clause 23. The method of clause 22, further comprising:

[0356] determining a format of the eradication treatment by selecting the format from a set of formats based on the at least one additional attribute.

[0357] Clause 24. The method of clause 22, wherein further comprising:

[0358] determining a format of the eradication treatment by creating the format based on the at least one additional attribute.

[0359] Clause 25. The method of clause 22, wherein the at least one additional attribute is an attribute of an unwanted plant.

[0360] Clause 26. The method of clause 25, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0361] Clause 27. The method of clause 22, wherein the at least one additional attribute is an attribute of the crop.

[0362] Clause 28. The method of clause 27, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0363] Clause 29. The method of clause 22, wherein the eradication treatment is a time-limited electrocution treatment and a format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant.

[0364] Clause 30. The method of any one of clauses 21-29, comprising:

[0365] processing the images to determine at least one additional location of non-contact in the images;

[0366] transforming each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; and

[0367] moving the electric treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

[0368] Clause 31. The method of any one of clauses 21-30, further comprising:

[0369] moving the platform amongst and through the crop; and

[0370] moving the electric treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

[0371] Clause 32. The method of any one of clauses 21-31, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

[0372] Clause 33. The method of clause 32, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0373] Clause 34. The method of any one of clauses 21-33, further comprising:

[0374] imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0375] Clause 35. The method of any one of clauses 21-34, further comprising:

[0376] training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

[0377] Clause 36. The method of any one of clauses 21-35, further comprising:

[0378] processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

[0379] Clause 37. The method of any one of clauses 21-36, further comprising:

[0380] transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of eradication treatments, wherein the electric treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0381] Clause 38. The method of any one of clauses 21-37, further comprising:

[0382] providing, by a depth-sensing subsystem supported by the platform, depth measurements; and

[0383] using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0384] Clause 39. The method of any one of clauses 21-38, wherein the actuation locations are coincident with the unwanted plants.

[0385] Clause 40. The method of any one of clauses 21-38, wherein the actuation locations are proximal to the unwanted plants.

[0386] Clause 41. A system for mechanically harming unwanted plants amongst a crop, the system comprising:

[0387] a platform positionable amongst the crop;

[0388] an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform;

[0389] a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;

[0390] a mechanical treatment subsystem supported by the platform, the mechanical treatment subsystem comprising an array of mechanical treatment modules extending across the platform, each of the mechanical treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually actuated to execute a mechanical harming action; and

[0391] a control subsystem operable to move each of the mechanical treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually actuate each of the mechanical treatment modules while at the respective actuation location thereby to execute the mechanical harming action to the unwanted plants.

[0392] Clause 42. The system of clause 41, wherein the processing subsystem processes the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem is operable to determine a format of the mechanical harming action based on the at least one additional attribute.

[0393] Clause 43. The system of clause 42, wherein the control subsystem is operable to determine a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

[0394] Clause 44. The system of clause 42, wherein control subsystem is operable to determine a format of the mechanical harming action by creating the format based on the at least one additional attribute.

[0395] Clause 45. The system of clause 42, wherein the at least one additional attribute is an attribute of an unwanted plant.

[0396] Clause 46. The system of clause 45, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0397] Clause 47. The system of clause 42, wherein the at least one additional attribute is an attribute of the crop.

[0398] Clause 48. The system of clause 47, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0399] Clause 49. The system of clause 42, wherein the format of the mechanical harming action comprises one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

[0400] Clause 50. The system of any one of clauses 41-49, wherein:

[0401] the processing subsystem processes the images received from the imaging subsystem to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart avoidance location in the three-dimensional region proximal to the platform; and

[0402] wherein the control subsystem is operable to move the at least one mechanical treatment module to one or more of the actuation locations in the three-dimensional region while avoiding coinciding with any avoidance location.

[0403] Clause 51. The system of any one of clauses 41-50, further comprising:

[0404] a drive subsystem for moving the platform amongst and through the crop, wherein the control subsystem moves the at least one mechanical treatment module to the actuation locations in coordination with moving of the platform by the drive subsystem.

[0405] Clause 52. The system of any one of clauses 41-51, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

[0406] Clause 53. The system of clause 52, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0407] Clause 54. The system of any one of clauses 41-53, wherein the imaging subsystem comprises an artificial lighting subsystem for imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0408] Clause 55. The system of any one of clauses 41-54, wherein the processing subsystem comprises a machine-learning model trained to segment unwanted plants in the images thereby to determine at least the locations of the unwanted plants.

[0409] Clause 56. The system of any one of clauses 41-55, wherein the processing subsystem is configured to process the images received from the imaging subsystem to determine at least locations in the images at which the mechanical harming action has already been delivered.

[0410] Clause 57. The system of any one of clauses 41-56, wherein the processing subsystem transforms the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions.

[0411] Clause 58. The system of any one of clauses 41-57, further comprising:

[0412] a depth-sensing subsystem supported by the platform and oriented to provide depth measurements for use by the processing subsystem to transform the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0413] Clause 59. The system of any one of clauses 41-58, wherein the actuation locations are coincident with the unwanted plants.

[0414] Clause 60. The system of any one of clauses 41-58, wherein the actuation locations are proximal to the unwanted plants.

[0415] Clause 61. A method for eradicating weeds amongst a crop, the method comprising:

[0416] positioning a platform amongst the crop;

[0417] capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform;

[0418] processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;

[0419] individually moving mechanical treatment modules that are supported by the platform in at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; and

[0420] individually actuating each of the mechanical treatment modules while at respective actuation locations thereby to deliver a mechanical harming action to the unwanted plants.

[0421] Clause 62. The method of clause 61, further comprising:

[0422] processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the mechanical harming action is based on the at least one additional attribute.

[0423] Clause 63. The method of clause 62, further comprising:

[0424] determining a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

[0425] Clause 64. The method of clause 62, further comprising:

[0426] determining a format of the mechanical harming action by creating the format based on the at least one additional attribute.

[0427] Clause 65. The method of clause 62, wherein the at least one additional attribute is an attribute of an unwanted plant.

[0428] Clause 66. The method of clause 65, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

[0429] Clause 67. The method of clause 62, wherein the at least one additional attribute is an attribute of the crop.

[0430] Clause 68. The method of clause 67, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

[0431] Clause 69. The method of clause 62, wherein format of the mechanical harming action comprises one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

[0432] Clause 70. The method of any one of clauses 61-69, comprising:

[0433] processing the images to determine at least one additional location of non-contact in the images;

[0434] transforming each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; and

[0435] moving the mechanical treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

[0436] Clause 71. The method of any one of clauses 61-70, further comprising:

[0437] moving the platform amongst and through the crop; and

[0438] moving the mechanical treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

[0439] Clause 72. The method of any one of clauses 61-71, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

[0440] Clause 73. The method of clause 72, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

[0441] Clause 74. The method of any one of clauses 61-73, further comprising:

[0442] imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

[0443] Clause 75. The method of any one of clauses 61-74, further comprising:

[0444] training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

[0445] Clause 76. The method of any one of clauses 61-75, further comprising:

[0446] processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

[0447] Clause 77. The method of any one of clauses 61-76, further comprising:

[0448] transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions, wherein the mechanical treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

[0449] Clause 78. The method of any one of clauses 61-77, further comprising:

[0450] providing, by a depth-sensing subsystem supported by the platform, depth measurements; and

[0451] using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

[0452] Clause 79. The method of any one of clauses 61-78, wherein the actuation locations are coincident with the unwanted plants.

[0453] Clause 80. The method of any one of clauses 61-78, wherein the actuation locations are proximal to the unwanted plants.

[0454] Clause 81. The system of any one of clauses 1-20, wherein each of the electric treatment modules of the array is fixed in an x-y direction with respect to the platform.

[0455] Clause 82. The system of any one of clauses 1-20, wherein at least one of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0456] Clause 83. The system of clause 82, wherein each of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0457] Clause 84. The system of any one of clauses 82 and 83, wherein the subregions of the electric treatment modules are non-overlapping.

[0458] Clause 85. The method of any one of clauses 21-40, wherein each of the electric treatment modules of the array is fixed in an x-y direction with respect to the platform.

[0459] Clause 86. The method of any one of clauses 21-40, wherein at least one of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0460] Clause 87. The method of clause 86, wherein each of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0461] Clause 88. The method of any one of clauses 86 and 87, wherein the subregions of the electric treatment modules are non-overlapping.

[0462] Clause 89. The system of any one of clauses 41-60, wherein each of the mechanical treatment modules of the array is fixed in an x-y direction with respect to the platform.

[0463] Clause 90. The system of any one of clauses 41-60, wherein at least one of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0464] Clause 91. The system of clause 90, wherein each of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0465] Clause 92. The system of any one of clauses 90 and 91, wherein the respective subregions of the mechanical treatment modules are non-overlapping.

[0466] Clause 93. The method of any one of clauses 61-80, wherein each of the mechanical treatment modules of the array is fixed in an x-y direction with respect to the platform.

[0467] Clause 94. The method of any one of clauses 61-80, wherein at least one of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0468] Clause 95. The method of clause 94, wherein each of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

[0469] Clause 96. The method of any one of clauses 94 and 95, wherein the subregions of the mechanical treatment modules are non-overlapping.

Claims

CLAIMSWhat is claimed is:

1. A system for electrocuting unwanted plants amongst a crop, the system comprising:a platform positionable amongst the crop;an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform;a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;an electric treatment subsystem supported by the platform, the electric treatment subsystem comprising an array of electric treatment modules extending across the platform, each of the electric treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually electrically actuated to deliver a time-limited electrocution treatment; and a control subsystem operable to move each of the electric treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually electrically actuate each of the electric treatment modules while at the respective actuation location thereby to deliver the time-limited electrocution treatment to the unwanted plants.

2. The system of claim 1, wherein the processing subsystem processes the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment based on the at least one additional attribute.

3. The system of claim 2, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment by selecting the format from a set of formats based on the at least one additional attribute.

4. The system of claim 2, wherein the control subsystem is operable to determine a format of the time-limited electrocution treatment by creating the format based on the at least one additional attribute.

5. The system of claim 2, wherein the at least one additional attribute is an attribute of an unwanted plant.

6. The system of claim 5, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

7. The system of claim 2, wherein the at least one additional attribute is an attribute of the crop.

8. The system of claim 7, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

9. The system of claim 2, wherein the format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant.

10. The system of any one of claims 1-9, wherein:the processing subsystem processes the images received from the imaging subsystem to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform, andwherein the control subsystem is operable to move the electric treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

11. The system of any one of claims 1-10, further comprising:a drive subsystem for moving the platform amongst and through the crop, wherein the control subsystem moves the electric treatment modules between the respective avoidance and actuation locations in coordination with moving of the platform by the drive subsystem.

12. The system of any one of claims 1-11, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

13. The system of claim 12, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

14. The system of any one of claims 1-13, wherein the imaging subsystem comprises an artificial lighting subsystem for imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

15. The system of any one of claims 1-14, wherein the processing subsystem comprises a machinelearning model trained to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

16. The system of any one of claims 1-15, wherein the processing subsystem is configured to process the images received from the imaging subsystem to determine at least locations in the images at which the time-limited electrocution treatment has already been delivered.

17. The system of any one of claims 1-16, wherein the processing subsystem transforms the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of time-limited electrocution treatments, wherein the control subsystem moves each of the electric treatment modules to a respective avoidance or actuation location based on the three-dimensional eradication map.

18. The system of any one of claims 1-17, further comprising:a depth-sensing subsystem supported by the platform and oriented to provide depth measurements for use by the processing subsystem to transform the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

19. The system of any one of claims 1-18, wherein the actuation locations are coincident with the unwanted plants.

20. The system of any one of claims 1-18, wherein the actuation locations are proximal to the unwanted plants.

21. A method for eradicating weeds amongst a crop, the method comprising:positioning a platform amongst the crop;capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform;processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;individually moving electric treatment modules that are supported by the platform along at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; andindividually actuating each of the electric treatment modules while at respective actuation locations thereby to deliver an eradication treatment to the unwanted plants.

22. The method of claim 21, further comprising:processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the eradication treatment is based on the at least one additional attribute.

23. The method of claim 22, further comprising:determining a format of the eradication treatment by selecting the format from a set of formats based on the at least one additional attribute.

24. The method of claim 22, wherein further comprising:determining a format of the eradication treatment by creating the format based on the at least one additional attribute.

25. The method of claim 22, wherein the at least one additional attribute is an attribute of an unwanted plant.

26. The method of claim 25, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

27. The method of claim 22, wherein the at least one additional attribute is an attribute of the crop.

28. The method of claim 27, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

29. The method of claim 22, wherein the eradication treatment is a time-limited electrocution treatment and a format of the time-limited electrocution treatment comprises one or more parameter selected from the group consisting of: a duration of electrocution, a power of electrocution, a pattern of power delivery of electrocution, a position of contact on an unwanted plant.

30. The method of any one of claims 21-29, comprising:processing the images to determine at least one additional location of non-contact in the images; transforming each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; andmoving the electric treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

31. The method of any one of claims 21-30, further comprising:moving the platform amongst and through the crop; andmoving the electric treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

32. The method of any one of claims 21-31, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

33. The method of claim 32, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

34. The method of any one of claims 21-33, further comprising:imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

35. The method of any one of claims 21-34, further comprising:training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

36. The method of any one of claims 21-35, further comprising:processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

37. The method of any one of claims 21-36, further comprising:transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least onethree-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of eradication treatments, wherein the electric treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

38. The method of any one of claims 21-37, further comprising:providing, by a depth-sensing subsystem supported by the platform, depth measurements; and using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

39. The method of any one of claims 21-38, wherein the actuation locations are coincident with the unwanted plants.

40. The method of any one of claims 21-38, wherein the actuation locations are proximal to the unwanted plants.

41. A system for mechanically harming unwanted plants amongst a crop, the system comprising:a platform positionable amongst the crop;an imaging subsystem supported by the platform and oriented to capture images of at least a portion of the crop that is proximal to the platform;a processing subsystem processing the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;a mechanical treatment subsystem supported by the platform, the mechanical treatment subsystem comprising an array of mechanical treatment modules extending across the platform, each of the mechanical treatment modules in the array individually controllable to be moved in at least a z-direction with respect to the platform between respective avoidance and actuation locations within the three-dimensional region and to be individually actuated to execute a mechanical harming action; and a control subsystem operable to move each of the mechanical treatment modules in the array along at least the z-direction between the respective avoidance and actuation locations in the three-dimensional region thereby to, while in an avoidance location, remain distal from a wanted plant and, while in an actuation location, contact or be proximal to an unwanted plant, the control subsystem operable to individually actuate each of the mechanical treatment modules while at the respective actuation location thereby to execute the mechanical harming action to the unwanted plants.

42. The system of claim 41, wherein the processing subsystem processes the images received from the imaging subsystem to determine at least one additional attribute, wherein the control subsystem is operable to determine a format of the mechanical harming action based on the at least one additional attribute.

43. The system of claim 42, wherein the control subsystem is operable to determine a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

44. The system of claim 42, wherein control subsystem is operable to determine a format of the mechanical harming action by creating the format based on the at least one additional attribute.

45. The system of claim 42, wherein the at least one additional attribute is an attribute of an unwanted plant.

46. The system of claim 45, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

47. The system of claim 42, wherein the at least one additional attribute is an attribute of the crop.

48. The system of claim 47, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

49. The system of claim 42, wherein the format of the mechanical harming action comprises one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

50. The system of any one of claims 41-49, wherein:the processing subsystem processes the images received from the imaging subsystem to determine at least one additional location of non-contact in the images and to transform each additional location of non-contact in the images into a counterpart avoidance location in the three-dimensional region proximal to the platform; andwherein the control subsystem is operable to move the at least one mechanical treatment module to one or more of the actuation locations in the three-dimensional region while avoiding coinciding with any avoidance location.

51. The system of any one of claims 41-50, further comprising:a drive subsystem for moving the platform amongst and through the crop, wherein the control subsystem moves the at least one mechanical treatment module to the actuation locations in coordination with moving of the platform by the drive subsystem.

52. The system of any one of claims 41-51, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

53. The system of claim 52, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

54. The system of any one of claims 41-53, wherein the imaging subsystem comprises an artificial lighting subsystem for imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

55. The system of any one of claims 41-54, wherein the processing subsystem comprises a machine-learning model trained to segment unwanted plants in the images thereby to determine at least the locations of the unwanted plants.

56. The system of any one of claims 41-55, wherein the processing subsystem is configured to process the images received from the imaging subsystem to determine at least locations in the images at which the mechanical harming action has already been delivered.

57. The system of any one of claims 41-56, wherein the processing subsystem transforms the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions.

58. The system of any one of claims 41-57, further comprising:a depth-sensing subsystem supported by the platform and oriented to provide depth measurements for use by the processing subsystem to transform the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

59. The system of any one of claims 41-58, wherein the actuation locations are coincident with the unwanted plants.

60. The system of any one of claims 41-58, wherein the actuation locations are proximal to the unwanted plants.

61. A method for eradicating weeds amongst a crop, the method comprising:positioning a platform amongst the crop;capturing, using an imaging subsystem supported by the platform, images of at least a portion of the crop that is proximal to the platform;processing, by a processing subsystem, the images received from the imaging subsystem to determine at least locations of wanted and unwanted plants in the images and to transform the locations in the images, respectively, into avoidance and actuation locations in a three-dimensional region proximal to the platform;individually moving mechanical treatment modules that are supported by the platform in at least a z-direction with respect to the platform between respective avoidance and actuation locations in the three-dimensional region thereby to, while in a respective avoidance location, remain distal from a wanted plant and, while in a respective actuation location, contact or be proximal to an unwanted plant; andindividually actuating each of the mechanical treatment modules while at respective actuation locations thereby to deliver a mechanical harming action to the unwanted plants.

62. The method of claim 61, further comprising:processing the images received from the imaging subsystem to determine at least one additional attribute, wherein a format of the mechanical harming action is based on the at least one additional attribute.

63. The method of claim 62, further comprising:determining a format of the mechanical harming action by selecting the format from a set of formats based on the at least one additional attribute.

64. The method of claim 62, further comprising:determining a format of the mechanical harming action by creating the format based on the at least one additional attribute.

65. The method of claim 62, wherein the at least one additional attribute is an attribute of an unwanted plant.

66. The method of claim 65, wherein the at least one additional attribute is selected from the group consisting of: moisture level of an unwanted plant, size of an unwanted plant, age of an unwanted plant, species of an unwanted plant.

67. The method of claim 62, wherein the at least one additional attribute is an attribute of the crop.

68. The method of claim 67, wherein the at least one additional attribute is selected from the group consisting of: moisture level of the crop, size of crop plants, age of the crop, species of the crop.

69. The method of claim 62, wherein format of the mechanical harming action comprises one or more parameter selected from the group consisting of: a strength of mechanical grip, a speed of mechanical pulling, a pattern of mechanical pulling, a position of contact on an unwanted plant.

70. The method of any one of claims 61-69, comprising:processing the images to determine at least one additional location of non-contact in the images; transforming each additional location of non-contact in the images into a counterpart additional avoidance location in the three-dimensional region proximal to the platform; andmoving the mechanical treatment modules to respective actuation locations in the three-dimensional region while avoiding coinciding with any additional avoidance location.

71. The method of any one of claims 61-70, further comprising:moving the platform amongst and through the crop; andmoving the mechanical treatment modules between respective avoidance and actuation locations in coordination with moving of the platform.

72. The method of any one of claims 61-71, wherein the imaging subsystem comprises at least one image capture device configured to capture respective images.

73. The method of claim 72, wherein the at least one image capture device is a video capture device and the respective images are frames of a digital video captured by the video capture device.

74. The method of any one of claims 61-73, further comprising:imparting artificial light to at least the portion of the crop imaged by the imaging subsystem.

75. The method of any one of claims 61-74, further comprising:training a machine-learning model to segment wanted and unwanted plants in the images thereby to determine at least the locations of the wanted plants and the locations of the unwanted plants.

76. The method of any one of claims 61-75, further comprising:processing the images received from the imaging subsystem to determine at least locations in the images at which treatment has already been delivered.

77. The method of any one of claims 61-76, further comprising:transforming the locations of wanted and unwanted plants in the images into avoidance and actuation locations in the three-dimensional region proximal to the platform by generating at least one three-dimensional eradication map comprising the avoidance and actuation locations in the three-dimensional region for a set of mechanical harming actions, wherein the mechanical treatment modules are each moved to a respective avoidance or actuation location based on the three-dimensional eradication map.

78. The method of any one of claims 61-77, further comprising:providing, by a depth-sensing subsystem supported by the platform, depth measurements; and using the depth measurements during the transforming of the locations of wanted and unwanted plants in the images captured by the imaging subsystem into avoidance or actuation locations in the three-dimensional region proximal to the platform.

79. The method of any one of claims 61-78, wherein the actuation locations are coincident with the unwanted plants.

80. The method of any one of claims 61-78, wherein the actuation locations are proximal to the unwanted plants.

81. The system of any one of claims 1-20, wherein each of the electric treatment modules of the array is fixed in an x-y direction with respect to the platform.

82. The system of any one of claims 1-20, wherein at least one of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

83. The system of claim 82, wherein each of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

84. The system of any one of claims 82 and 83, wherein the subregions of the electric treatment modules are non-overlapping.

85. The method of any one of claims 21-40, wherein each of the electric treatment modules of the array is fixed in an x-y direction with respect to the platform.

86. The method of any one of claims 21-40, wherein at least one of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

87. The method of claim 86, wherein each of the electric treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

88. The method of any one of claims 86 and 87, wherein the subregions of the electric treatment modules are non-overlapping.

89. The system of any one of claims 41-60, wherein each of the mechanical treatment modules of the array is fixed in an x-y direction with respect to the platform.

90. The system of any one of claims 41-60, wherein at least one of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

91. The system of claim 90, wherein each of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

92. The system of any one of claims 90 and 91, wherein the respective subregions of the mechanical treatment modules are non-overlapping.

93. The method of any one of claims 61-80, wherein each of the mechanical treatment modules of the array is fixed in an x-y direction with respect to the platform.

94. The method of any one of claims 61-80, wherein at least one of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

95. The method of claim 94, wherein each of the mechanical treatment modules of the array is moveable within a respective subregion in one or more of the x and y directions with respect to the platform.

96. The method of any one of claims 94 and 95, wherein the subregions of the mechanical treatment modules are non-overlapping.

Images