Carrying out and documenting the application of crop protection products

Abstract

The present invention relates to the technical field of applying crop protection products. The present invention relates to a device and to a method for automated and precise application of one or more crop protection products to a target object and for documenting the application.

Description

[0001] CARRYING OUT AND DOCUMENTING AN APPLICATION OF PEST PROTECTION PRODUCTS

[0002] Technical area

[0003] The present invention relates to the technical field of plant protection product application. The present invention relates to a device and a method for the automated and precise application of one or more plant protection products to a target object and for documenting the application.

[0004] Introduction

[0005] Plant protection products are used worldwide to protect plants or plant products from pests or to prevent their action, to destroy undesirable plants or parts of plants, to inhibit undesirable plant growth or to prevent such growth, and / or to influence the life processes of plants in a way other than nutrients (e.g. growth regulators).

[0006] Autonomous systems are increasingly being used that automatically detect crops and / or unwanted companion plants and apply pesticides in a targeted manner.

[0007] WO2017 / 178666A1, for example, discloses a set of autonomous devices that use a camera to capture images of a crop field and feed the resulting images to an artificial neural network trained to distinguish wild plants from cultivated plants. The information about the specific plant is used to apply crop protection products in a targeted manner.

[0008] WO2021 / 144785A1 discloses a system for the biological real-time detection and treatment of a predefined plant condition, comprising a control system that recognizes predefined phenotypic characteristics of plants based on image recordings, and an execution unit configured to treat the predefined plant condition. The identification of plants and / or plant conditions is accompanied by a targeted treatment of individual plants.

[0009] US2021 / 090274 describes an agricultural device with image sensors. While the device moves through a field, images are captured and analyzed. The distance between the sensor and the object, each represented by a pixel, as well as the crop, is determined, which then triggers the treatment of the plant.

[0010] CN107128484 describes how an autonomously flown aircraft equipped with inkjet technology sprays an insecticide at low altitude, monitored by cameras. However, the systems currently on the market do not guarantee that the pesticide will only reach the intended target area, avoiding surrounding areas, and thus in the smallest amount that still achieves the desired effect.

[0011] In organic farming, efforts are made to avoid pesticides as much as possible. The prevention of diseases and pests is paramount. All cultivation measures, such as crop rotation, fertilization, soil cultivation, and species and variety selection, are aimed at limiting the spread of pests.

[0012] If diseases or pests occur despite these preventive measures, only a limited range of plant protection products can be used in organic farming. The use of plant protection products is regulated by EU legislation on organic farming: "If a threat to crops is identified, only plant protection products approved for use in organic production in accordance with Article 16 may be used" (Article 12 of Regulation 834 / 2007).

[0013] With Regulation 834 / 2007, the EU aims to minimize the entry of plant protection products into the environment to avoid unwanted consequences. The regulation is based on the belief that the application of plant protection products under real field conditions inevitably leads to plant protection product contamination of the environment.

[0014] Based on the described state of the art, the task is to design the use of plant protection products in such a way that the plant protection product is verifiably applied in the smallest possible quantity only where it is intended.

[0015] Summary of the present invention

[0016] The present invention provides means with which, on the one hand, plant protection products can be applied to a target object in a targeted manner and in the smallest possible amount in which the plant protection product can still exert its effect, and, on the other hand, it is demonstrated that the use of plant protection products is limited to the target object.

[0017] A first object of the present invention is a method comprising the steps:

[0018] Creating a first image of a target object,

[0019] Identifying the target object based on the first image capture,

[0020] Determination of the required application quantity based on the properties of the target object. Determination of a target region based on the first image and / or on the basis of a second image, Applying an amount of a plant protection product to the target region using a dosing unit,

[0021] Creating a third image, wherein the third image shows the plant protection product applied to the target region,

[0022] Creating a data set comprising o the first and / or the second image recording, o information about the identified target object, and o the third image recording,

[0023] Outputting and / or storing the data set and / or transmitting the data set to a computer system.

[0024] A further subject of the present invention is a device comprising a control and computing unit, at least one image recording unit, a dosing unit, and a data storage and / or a transmission unit, wherein the control and computing unit is configured to receive a first image from the at least one image recording unit, to identify a target object in the first image recording, optionally to receive a second image recording from the at least one image recording unit, to determine a target region based on the first and / or second image recording, to cause the dosing unit to apply an amount of a plant protection agent to the target region, to cause the at least one image recording unit to generate a third image recording, wherein the third image recording shows the plant protection agent applied to the target region, to generate a data set, wherein the data set contains the first and / or the second image recording,o information about the identified target object, and o the second image capture, and storing the data set in the data memory and / or causing the transmitting unit to transmit the data set to a computer system. Detailed description of the invention,

[0025] The invention is explained in more detail below, without distinguishing between the subject matters (method, device). Rather, the following explanations are intended to apply analogously to all subject matters of the invention, regardless of the context (method, device) in which they occur.

[0026] If steps are mentioned in a particular order in this description or in the claims, this does not necessarily mean that the invention is limited to that order. Rather, it is conceivable that the steps may be performed in a different order or even in parallel; unless a step builds on another step, which absolutely requires that the subsequent step be performed (which will become clear in individual cases). The specified sequences thus represent preferred embodiments of the invention.

[0027] The present invention provides means for the targeted treatment of a target object with a plant protection agent.

[0028] The term "plant protection product" refers to a product used to protect plants or plant products from harmful organisms or to prevent their effects, to destroy undesirable plants or parts of plants, to inhibit or prevent undesirable plant growth, and / or to influence plant life processes in a manner other than by providing nutrients (e.g., growth regulators). Examples of plant protection products include herbicides, fungicides, and insecticides. The plant protection products mentioned above are also referred to as pesticides.

[0029] A plant protection agent within the meaning of the present invention is preferably a plant protection agent that can be converted into a liquid form. It is preferably liquid in a temperature range of 4°C to 35°C under a pressure of 1,013 mbar (atmospheric pressure at sea level). However, it is also conceivable for it to be in a different state of aggregation and to be converted into a liquid form only before application (e.g., by heating in the case of a solid plant protection agent or by condensing a gaseous plant protection agent).

[0030] "Liquid" means that the pesticide can be transformed into a droplet by an external force. This also includes pasty media that fall under the term liquid.

[0031] It is conceivable that several (different) plant protection products are applied simultaneously or sequentially. For the sake of clarity, this description primarily refers to one plant protection product. However, this is not to be understood as a limitation to a single plant protection product; rather, the term "plant protection product" is intended to encompass multiple plant protection products as well as a combination of plant protection products with other substances and / or substance mixtures, such as plant nutrients.

[0032] The target object to be treated with a pesticide may be a pest.

[0033] A "pest" is defined as an organism that can appear during the cultivation of crops and damage the crops, negatively impact the crop's yield, or compete with the crop for natural resources. Examples of such pests include weeds, grass weeds, animal pests such as beetles, caterpillars, and worms, fungi, and pathogens (e.g., bacteria and viruses). Even though viruses are not considered organisms from a biological perspective, they are nevertheless considered to be pests for the purposes of this paper.

[0034] In a preferred embodiment, the target object is a weed plant. The term “weed” refers to plants that spontaneously accompany vegetation in crop stands, grassland, or gardens that are not deliberately cultivated there and that develop, for example, from the seed potential of the soil or via airborne migration. The term is not restricted to herbs in the true sense of the word, but also includes grasses, ferns, mosses, and woody plants. In the field of plant protection, the term “weed” is often used to clarify the distinction from herbaceous plants. In this text, the term “weed” is used as a generic term to encompass the term “weed grass.” For the purposes of the present invention, weeds are therefore plants that accompany the cultivation of a desired crop.Since they compete with the crop for resources, they are undesirable and should therefore be controlled.

[0035] In a further preferred embodiment, the target object is an animal pest, such as an arthropod. Approximately two-thirds of currently known animal species belong to the arthropods (phylum Arthropoda), 85% of which are insects. A significant proportion of arthropods are phytophagous: these animals feed on plants and can impair growth, cause sucking and feeding damage, and transmit viral diseases. This causes, for example, significant yield and quality losses in the cultivation of crops. In a preferred embodiment, the animal pest is an insect (in the various stages from larva (caterpillar, dewlap) to adult stage) or an arachnid.Particularly preferably, the pest is an agricultural pest such as codling moth, aphid, natterjack, fruit skin tortrix, Colorado potato beetle, cherry fruit fly, cockchafer, European corn borer, plum tortrix, rhododendron leafhopper, seed moth, scale insect, gypsy moth, spider mite, grape tortrix, walnut fruit fly, whitefly, large rapeseed stem weevil, spotted cabbage stem weevil, rapeseed pollinator, cabbage pod weevil, cabbage pod midge or rapeseed flea, or a forest pest such as aphid, blue pine jewel beetle, bark beetle, oak jewel beetle, oak processionary moth, oak tortrix, spruce web leaf wasp, common woodworm, large brown barkeater, pine bush wasp, pine moth, pine moth, pine moth, small Spruce leaf wasp, nun moth, horse chestnut leaf miner, gypsy moth or sapwood beetle.

[0036] In a further preferred embodiment, the target object is a cultivated plant infected by a fungus, a bacterium, or a virus. The term "cultivated plant" refers to a plant that is purposefully cultivated as a useful or ornamental plant through human intervention.

[0037] The treatment of a target object according to the invention can be carried out in four steps:

[0038] 1. Identify the target object and determine the dosage

[0039] 2. Determine a target region

[0040] 3. Applying a pesticide to the target area

[0041] 4. Documenting the application

[0042] These steps are carried out automatically, i.e. without human intervention.

[0043] These steps, which in turn may include several substeps, are described in more detail below.

[0044] In a further preferred embodiment, a method is described, wherein in step (c) the determination of the target region is additionally carried out on the basis of the first and second image recordings, which are respectively generated from different directions, and wherein in step (f) the data set additionally comprises the second image recording and depth information relating to a distance of the dosing unit from the target region, which was obtained from the two image recordings.

[0045] In a further preferred embodiment, a method is described wherein, in step (d), the plant protection agent is additionally applied in the form of one or more individual drops.

[0046] In a further preferred embodiment, a method is described, wherein additionally in step (e) the third image recording shows one or more individual drops that have been applied to the target region in step (d).

[0047] In a further preferred embodiment, a method is described, wherein the data set generated in step (f) additionally comprises the total amount of the plant protection agent applied in the target region.

[0048] In a further preferred embodiment, a method is described wherein, additionally based on the data set generated in step (f) and the third image contained therein, it is determined whether and what quantity of crop protection agent has arrived in the target region. In a further preferred embodiment, a method is described wherein, additionally in step (d), the dosing unit is installed on a windbreak, which ensures that one or more individual drops of crop protection agent do not drift from the desired target trajectory on their trajectory from the dosing unit toward the target region.

[0049] In a further preferred embodiment, a device comprising a control and computing unit, at least one image recording unit, a dosing unit, and a data storage and / or a transmission unit is described, wherein the control and computing unit is configured to receive a first image from the at least one image recording unit, to identify a target object in the first image recording, optionally to receive a second image recording from the at least one image recording unit, to determine a target region based on the first and / or second image recording, to cause the dosing unit to apply an amount of a plant protection agent to the target region, to cause the at least one first image recording unit to generate a third image recording, wherein the third image recording shows the plant protection agent applied to the target region, to generate a data set, wherein the data set comprises the first and / or the second image recording,information about the detected target object, and the second image recording, and storing the data set in the data memory and / or causing the transmitting unit to transmit the data set to a computer system.,

[0050] In a further preferred embodiment, a device is described, wherein the control and computing unit is configured to receive a second image from a second image recording unit, wherein the second image recording unit is spatially positioned differently relative to the first image recording unit, to determine a target region on the basis of the first and second image recordings, to cause the dosing unit to apply an amount of a crop protection agent to the target region, to cause the at least one first image recording unit to generate a third image recording, wherein the third image recording shows the crop protection agent applied to the target region, to generate a data set, wherein the data set comprises the first image recording, information about the determined target object, and the second image recording.

[0051] In a further preferred embodiment, a device is described, wherein the dosing unit contains a storage container, a nozzle, a precision valve and a microreservoir which is suitable for applying drops with a volume of 1 to 100 nL.

[0052] In a further preferred embodiment, a device is described, wherein the drop has a volume of 1 to 50 nL.

[0053] In a further preferred embodiment, a device is described, wherein the first image recording unit and the second image recording unit are designed in the form of a stereo camera. In a further preferred embodiment, a device is described, wherein the dosing unit is installed on a windshield, which ensures that individual drops of the crop protection agent do not drift from the desired target trajectory on their trajectory from the dosing unit toward the target region.

[0054] Identifying the target object

[0055] Before a target object can be treated with a pesticide, the target object must first be detected (identified).

[0056] According to the invention, the target object is identified based on an image recording showing the target object. It is conceivable that the identification is performed based on multiple image recordings (e.g., recorded at different time intervals and / or from different directions (viewing angles)). The one image recording or the multiple image recordings used to identify the target object are also referred to in this description as the first image recording(s).

[0057] The term "image" preferably refers to a two-dimensional representation of an object or part of it. The image is typically a digital image. The term "digital" means that the image can be processed by a machine, usually a computer system. "Processing" refers to the well-known methods of electronic data processing (EDP). Digital images can be processed, edited, and reproduced using computer systems and software, and can also be converted into standardized data formats such as JPEG (a graphics format of the Joint Photographie Experts Group), PNG (Portable Network Graphics), or SVG (Scalable Vector Graphics). Digital images can be visualized using suitable display devices, such as computer monitors, projectors, and / or printers.

[0058] One or more initial images of the target object can be generated using one or more image acquisition units. Such an image acquisition unit can be a camera, for example.

[0059] A camera typically comprises an image sensor and optical elements. The image sensor is a device for capturing two-dimensional images from light electrically. It is typically a semiconductor-based image sensor, such as a CCD (charge-coupled device) or CMOS (complementary metal-oxide-semiconductor) sensor. The optical elements (lenses, apertures, and the like) serve to create the sharpest possible image of the object, of which a digital image is to be captured, on the image sensor.Other devices and systems that can produce an image of an object, such as an array of hyperspectral sensors, a thermal imaging camera (infrared camera), a laser scanner, a LIDAR system (LIDAR = light detection and ranging or light imaging, detection and ranging), an ultrasound source in combination with an ultrasound sensor and / or combinations of several units, are also intended to fall under the term “image recording unit”.

[0060] If the generation of an image is based on the detection of the spatial distribution of electromagnetic rays, the electromagnetic rays can be part of the sunlight and / or be generated by one or more artificial sources of electromagnetic radiation, such as LED (light-emitting diode) and / or halogen lamps.

[0061] To identify a target object based on one or more first image recordings showing at least part of the target object, a trained machine learning model is preferably used.

[0062] A "machine learning model" can be understood as a computer-implemented data processing architecture. The model can receive input data and produce output data based on this input data and model parameters. Through training, the model can learn a relationship between the input data and the output data. During training, the model parameters can be adjusted to produce a desired output for a given input.

[0063] When training such a model, the model is presented with training data from which it can learn. The trained machine learning model is the result of the training process. The training data includes input data and the correct output data (target data) that the model is supposed to generate based on the input data. During training, patterns are recognized that map the input data to the target data.

[0064] During the training process, the input data of the training data is fed into the model, and the model generates output data. The output data is compared with the target data. Model parameters are adjusted to reduce the deviations between the output data and the target data to a (defined) minimum.

[0065] During training, an error function (loss function) can be used to evaluate the model's prediction quality. The error function can be chosen to reward a desired relationship between output and target data and / or penalize an undesirable relationship between output and target data. Such a relationship can be, for example, a similarity, a dissimilarity, or some other relationship. An error function can be used to calculate an error value (loss value) for a given pair of output and target data. The goal of the training process can be to modify (adjust) the parameters of the machine learning model so that the error value is reduced to a (defined) minimum for all pairs in the training dataset.

[0066] For example, an error function can quantify the deviation between the model's output data for a given input and the target data. For example, if the output and target data are numbers, the error function can be the absolute difference between these numbers. In this case, a high absolute value of the error function may indicate that one or more model parameters need to be significantly changed.

[0067] For example, for output data in the form of vectors, difference metrics between vectors such as the mean square error, a cosine distance, a norm of the difference vector such as a Euclidean distance, a Chebyshev distance, an Lp norm of a difference vector, a weighted norm, or any other type of difference metric between two vectors can be chosen as the error function.

[0068] For higher-dimensional outputs, such as two-dimensional, three-dimensional, or higher-dimensional outputs, an element-wise difference metric can be used. Alternatively or additionally, the output data can be transformed, e.g., into a one-dimensional vector, before calculating a loss value.

[0069] In the present case, the machine learning model can be trained to output information based on one or more first image recordings that at least partially show a target object, wherein the information indicates which target object is involved. The machine learning model can be configured as a classification model that assigns one of several classes to one or more first image recordings as input data, wherein each class represents a defined target object.

[0070] For example, the machine learning model can be trained to distinguish weeds from crops. In such a case, the classification model can be configured to assign one or more initial image captures to one of two classes, where a first class represents weeds and the other class represents plants (or other objects) that are not weeds (e.g., crops).

[0071] For example, the machine learning model can be trained to recognize specific weeds and / or specific crops. Each specific plant can be represented by a class.

[0072] Likewise, the machine learning model can be trained to recognize (specific or non-specific) pests, such as (specific or non-specific) insects and / or arachnids. The term "specific" means that a defined phylum, class, order, family, genus, and / or species of pest is considered. The term "non-specific" means that only a distinction is made between pests and non-pests, weeds and non-weeds, animal pests and animal non-pests, or the like.

[0073] Likewise, the machine-based harvesting model can be trained to distinguish between animal pests and beneficial organisms. A beneficial organism is defined as an organism for which a pest serves as a food source or host, or which is important for a successful harvest for other reasons (e.g., as a pollinator).

[0074] Likewise, the machine learning model can be trained to distinguish a crop infected by a pathogen (e.g., bacteria, virus) and / or a fungus from a healthy, non-infected crop. The machine learning model can also be trained to recognize specific pathogens and / or fungi.

[0075] It is also conceivable that the machine learning model is trained to recognize two objects (a first and a second object) simultaneously or sequentially, or that there are multiple machine learning models, a first model being trained to recognize a first object and a second model being trained to recognize a second object. The first object can be, for example, a (specific) crop, and the second object can be, for example, a (specific) pest for the (specific) crop.

[0076] A machine learning model can be trained based on training data. The training data can include a large number of reference images of target objects and information about which target object each is. The term "large number" preferably means more than 100. The term "reference image" merely serves to distinguish between images used to train a machine learning model and images used when the trained machine learning model is used to recognize a target object in an image. The information about which target object each is in an image can, for example, be an identification number assigned to the target object (target value).

[0077] The reference images can be fed to the machine learning model, and the machine learning model can be configured to generate an output that identifies the target object, such as an identification number. The output can be compared with the target value. Using an error function, the deviation between the output and the target value can be quantified. Parameters of the machine learning model can be modified to reduce the deviation to a (defined) minimum. This can be done, for example, using a backpropagation method.

[0078] If the deviation between the output and the target value is reduced to a (defined) minimum for a large number of reference images, the trained machine learning model can be used to detect target objects in unknown images. The term "unknown" means that a corresponding image was not used in training the model.

[0079] Preferably, the machine learning model is an artificial neural network or the model includes one.

[0080] An artificial neural network comprises at least three layers of processing elements: a first layer with input neurons (nodes), an Nth layer with at least one output neuron (node), and N-2 inner layers, where N is a natural number and greater than 2.

[0081] The input neurons receive one or more image recordings. The output neurons provide output data that characterizes a target object shown in an image recording.

[0082] The processing elements of the layers between the input neurons and the output neurons are connected in a predetermined pattern with predetermined connection weights.

[0083] The neural network can be trained, for example, using a backpropagation method. The goal is to achieve the most reliable mapping possible from given input data to given output data. The quality of the mapping is described by an error function. The goal is to minimize the error function. With the backpropagation method, an artificial neural network is trained by changing the connection weights.

[0084] In the trained state, the connection weights between the processing elements contain information regarding the relationship between image captures and transformed image captures.

[0085] A cross-validation method can be used to split the data into training and validation sets. The training set is used for backpropagation training of the network weights. The validation set is used to test the prediction accuracy of the trained network when applied to unknown data. Machine learning models for pest detection are described in the prior art (see, for example, W02021090294A1, DE102018222428A1, US20200045953A1,

[0086] WO2019 / 046203A1, EP3804488A1, US20210383535, US10599959).

[0087] Determine a target region

[0088] Once a target object has been identified, the next step is to determine a target region to which one or more plant protection products are to be applied. The target region can be the target object or a part of it. In the case of an animal pest, the target region is preferably the target object itself / as a whole or a body part such as the head. A target region can also be target objects that represent certain reproductive stages of an animal pest, such as larvae or eggs. In the case of phytopathogenic fungi, the target region is the area of ​​the crop plant that is infested by the fungus.

[0089] In the case of a weed, the target region is preferably a part of the target plant that is particularly sensitive to pesticides and is of great importance for the growth or reproduction of the target plant. Examples include stems, roots, leaf axils, shoot axils, shoot roots, leaf roots, cotelydons, buds, flowers, or parts of flowers.

[0090] If the target is a pest, the region where a pesticide will produce an effect (preferably an optimal effect) is preferably selected as the target region. For example, if the target is a weed, and the pesticide to be applied (e.g., a herbicide) is absorbed by the weed through its leaves and then causes damage to the weed, a leaf is selected as the target region. A side of a leaf can also be selected as the target region, for example, the upper side of a leaf (the side of a leaf facing the sun) and / or the underside of a leaf (the side of a leaf facing away from the sun).

[0091] It is also conceivable that several target regions are selected, e.g. in the case of a weed plant, several leaves and / or several areas on one leaf / on several leaves.

[0092] With the application technique described in this specification, drops of a liquid pesticide with a volume of less than 15 nL can be generated. A spherical drop of this size has a diameter of approximately 0.3 mm. When applied to a target object, such a drop wets an area of ​​approximately 0.07 mm. 2 .

[0093] It is therefore conceivable that a target region is chosen that is significantly smaller than 1 mm 2Furthermore, it is conceivable to wet a larger area than the wetting area of ​​a single droplet by placing a large number of individual droplets next to one another in the target region in a grid pattern (e.g., in the form of a square or hexagonal dot grid). Depending on the implementation, the step "Determining a target region" can have multiple meanings and / or include multiple substeps. If the target region is identical to the target object, the target object can be selected as the target region.

[0094] If the target region is part of the target object, the step "Determining a target region" may include the substep "Identifying a target region." Identifying the target region is preferably based on one or more images of the target object.

[0095] One or more image recordings used to identify a target region are referred to in this description as a second image recording(s). It is conceivable that one or more second image recordings are identical to one or more first image recordings. In other words, it is conceivable that the image recording(s) used to identify the target object is / are also used to identify the target region; however, it is also conceivable that other and / or further image recordings are generated on the basis of which one or more target regions are identified.

[0096] Thus, in a first step, a weed plant can be identified based on a first image or several first image recordings (the target object), and in a second step, a leaf or a region of a leaf of the weed plant can be identified based on a second image recording or several second image recordings (the target region). The second image recording(s) can be completely or partially identical to the first image recording(s), or different from them.

[0097] One or more second images can be generated (like the first image(s)) by at least one image capture unit, e.g., a camera. Preferably, optical elements (e.g., a camera lens) are used that result in an enlarged image (enlarged compared to the original) of the target region in the image. A zoom lens is preferably used.

[0098] In a preferred embodiment, multiple image recording units are used, each generating second image recording of the target region from different directions. In one embodiment, the first and second image recordings are generated from different directions. In one embodiment, the first and second image recordings are generated from different directions using a first and a second image recording unit. In this way, it is possible to obtain depth information that provides information about distances. Such depth information can be used, for example, to determine the distance of a nozzle of the dosing unit from the target region and / or to position the nozzle of the dosing unit relative to the target region. A stereo camera can be used, for example.

[0099] Applying a plant protection product to the target area

[0100] Once a target region has been identified, the pesticide is applied to the target area. The application is carried out using a dosing unit. It is conceivable that multiple dosing units may be used. Such a dosing unit comprises at least one nozzle. The pesticide is applied to the target area from the nozzle.

[0101] The dosing unit also includes at least one dispenser for the precise dosing of liquids.

[0102] The dosing unit further comprises at least one precision valve that controls the discharge of individual drops through the nozzle. Preferably, a piezo actuator is used for dosing.

[0103] The dosing unit further comprises an orientation unit, for example, an XYZ traverse, which can move the dosing unit along three or more axes. The orientation unit can also rotate the dosing unit in a specific spatial direction to change its trajectory. Preferably, the trajectory is not vertical.

[0104] The dosing unit further comprises a storage container for holding the crop protection agent and a microreservoir into which a defined amount of crop protection agent flows from the storage container before this defined amount is applied via at least one nozzle. Crop protection agent can flow from the storage container into the microreservoir via a connection. This connection is preferably a pipe or hose.

[0105] The step of "applying a plant protection product to the target region" may include the step of "positioning the dosing unit relative to the target region." The application of a plant protection product may be vertical (relative to a surface of the target region), from above and / or from below, and / or at an angle other than 90°.

[0106] Preferably, the application is carried out at an angle of 45° to 90° with respect to a surface of the target region.

[0107] The dosing unit is preferably positioned at a defined distance from the target region with the aid of the orientation unit. The distance is preferably kept constant within defined limits: if a new target region is selected / determined, a defined distance within the defined limits is set between the dosing unit or at least one nozzle of the dosing unit and the target region. The distance between at least one nozzle and the target region is preferably at least 10 cm, preferably at least 20 cm. This preferably makes it possible to hit target points on the target region in a conical area below the dosing unit. The plant protection agent is preferably applied in the form of one or more individual drops whose volume is in the range from 1 nL to 100 nL, more preferably in the range from 1 nL to 50 nL, and even more preferably in the range from 1 to 30 nL.

[0108] It has been demonstrated that droplets of the specified size can be applied from a distance of 10 cm or more to the target region (e.g., the upper or lower surface of a leaf) of a target object (e.g., a crop infected with a fungus or pathogen, or a weed plant) without the droplets bouncing off the surface of the target region and / or breaking up into smaller droplets. The applied droplets of the specified size reach the surface and adhere there.

[0109] The application of the pesticide in the form of individual drops is preferably carried out using a dosing unit for contactless dosing. Contactless dosing—also known as jetting, jet process, and contactless dosing—is a method in which a metered, continuous amount of liquid flows freely—i.e., completely detached from the dosing opening—toward the target region. Contactless dosing allows for highly precise portioning of liquid and pasty media and precise application.

[0110] A dosing unit for contactless dosing can have a nozzle in which a movable ejection element, e.g. a plunger, can be arranged. To eject pesticide, the ejection element inside the nozzle can be pushed forward in an ejection direction at high speed towards a nozzle opening, whereby a single drop of pesticide is ejected from the nozzle. This process is generally and hereinafter referred to as the ejection process. The ejection element can then be retracted in an opposite retraction direction. The size of the droplets or the amount of pesticide per droplet can be predetermined by the design and control of the nozzle as well as by the resulting effect.

[0111] Preferably, the plant protection agent is "actively" ejected from the nozzle by an (ejection) movement of the ejection element relative to the nozzle. During the ejection process, an ejection tip of the ejection element preferably comes into contact with the plant protection agent to be dispensed and "pressurizes" or "pushes" the plant protection agent out of the nozzle of the dosing unit due to the (ejection) movement of the ejection element and / or the nozzle. The dosing unit according to the present invention thus differs from other dosing systems in which a movement of a closure element merely leads to an opening of the nozzle, with a pressurized dosing substance then exiting the nozzle of its own accord. This is the case, for example, with injection valves of internal combustion engines. Preferably, the ejection element can also be brought into a closed position by firmly connecting it to a sealing seat of the nozzle opening in the nozzle and temporarily remaining there.In the case of more viscous pesticides, it may also be sufficient for the ejection element to simply remain in the retracted position, i.e. away from the sealing seat, without a single drop of the pesticide escaping.

[0112] The movement of the ejection element required to eject the pesticide can be achieved with the help of an actuator unit of the dosing system. Such an actuator unit can, in principle, be implemented in various ways, e.g., using a pneumatically or hydraulically operated actuator. Piezoelectric and / or electromagnetically operated actuators are preferably used.

[0113] Alternatively or in addition to the movable ejection element, the nozzle of the dosing system itself can be moved in an ejection or retraction direction to dispense the pesticide. To dispense the pesticide, the nozzle and an ejection element arranged inside the nozzle can be moved toward or away from each other in a relative movement, whereby the relative movement can occur either solely through a movement of the nozzle or at least partially also through a corresponding movement of the ejection element.

[0114] After the ejection movement has ended, the micro reservoir can be automatically refilled from the storage container.

[0115] During dosing, the individual drops preferably reach a speed of 1 m / s to 100 m / s, very preferably 5 m / s to 50 m / s, particularly preferably 5 m / s to 25 m / s.

[0116] Dosing systems for contactless dosing of a dosing substance are described in the prior art (see, for example, WO2021 / 028197A1, WO2020 / 120176A2, WO2019 / 197181A1, WO2015 / 192896A1, W02014 / 140195A1); they are typically used for the application of soldering agents or adhesives.

[0117] As already described, a single drop of a plant protection product can be applied, or multiple drops can be applied. In the case of an animal pest, a single drop may be sufficient to render the pest harmless. In the case of a weed and / or a crop plant infested with a pathogen and / or fungus, it may be necessary to apply multiple drops to control the weed and / or pathogen and / or fungus. Multiple drops can be applied to the target region in the form of a defined pattern, e.g., in the form of a square or hexagonal dot matrix. A distribution of drops along defined characteristics of the target region is also conceivable. For example, multiple drops can be applied along the midvein of a leaf and / or along lateral veins of a leaf.In a preferred embodiment, the amount of crop protection agent to be applied is automatically determined based on the one or more second images showing the target region. "Amount" can refer to the volume of a single droplet and / or the number of individual drops to be applied. The spacing of the applied individual drops in the target region and / or the distribution of the individual drops in the target region can also be automatically determined / specified based on the one or more second images.

[0118] The high precision and small quantity of applied drops has the advantage that no pesticide is transferred from the target region to other regions, especially the soil, and thus only a minimal surface comes into contact with the pesticide.

[0119] In a preferred form, the dosing unit can be coupled to an autonomous, mobile robot platform, preferably weather-independent. Thus, in addition to being guided by the orientation unit, the dosing unit is moved in the direction of travel. However, other forms are also possible, e.g., an attachment, UAV, or handheld device.

[0120] The dosing unit can also be installed on a passive or active draft shield to prevent droplet drift. The active draft shield influences the airflow in the dispenser area with pneumatic backpressure. For this purpose, the local wind speed and gust factor are measured. A model-based prediction algorithm controls the pneumatic system. This algorithm considers the gust factor and / or the local wind speed to determine the pneumatic backpressure. The passive draft shield is characterized by a protective device, such as a protective shield, that shields the dosing unit.

[0121] Documenting the application

[0122] The "Documenting the Application" step serves to demonstrate that the application is limited to the target region. In other words, documenting the application ensures that the plant protection product has been applied exclusively to the target region and that no plant protection product has entered any area outside the target region.

[0123] The application is documented on the basis of a third image recording or on the basis of several third image recordings.

[0124] Such a third image shows the target region after the application of one or more individual droplets of the plant protection product. Such a third image shows one or more individual droplets in the target region. Preferably, the generation of an image of the target region is coordinated with the ejection movement of the ejection element such that an image of the target region is generated before an ejection movement and a defined period of time after an ejection movement. The defined period of time is preferably longer than the time required for an individual droplet to travel from the nozzle opening to the target region after the ejection element has ejected the individual droplet. In this way, an image of the target region can be generated before the application of an individual droplet and another image of the target region after the application of the individual droplet.By comparing the two images (before / after), it can be shown that in the image taken a defined time period after the ejection movement, a single droplet has been added to the target region compared to the image taken before the ejection movement. Coupling the ejection movement with the generation of images of the target region thus allows for verification that every single droplet ejected has actually reached the target region.

[0125] The one or more third image recordings are generated using at least one image recording unit; preferably, the same image recording unit is used that was also used to identify / determine the target region.

[0126] It is conceivable that the ejection process could also be documented, i.e., that one or more images could be taken to document the moment a single drop leaves the nozzle opening of the dosing device. A drop detection unit, such as that described in WO2017 / 060336A1, could be used for this purpose.

[0127] Preferably, the application process is recorded in the form of a data set. Such a data set preferably comprises one or more first images of the target object, information about which target object was identified based on the one or more first images, and one or more third images showing one or more applied individual drops.

[0128] The data set may further comprise one or more second image recordings.

[0129] The first one or more images show the target object being treated with pesticide.

[0130] The information about which target object was identified based on the one or more first images can be used to detect errors in the identification. The one or more first images of the target object can be viewed by a human trained to recognize the target object in the one or more first images. The human can then check the information about which target object was identified in the one or more first images by the device according to the invention and determine whether the identification by the device according to the invention was correct or incorrect.

[0131] The one or more second images provide information about which target region has been selected / determined for the application of the plant protection product.

[0132] The one or more third images provide information about whether and what quantity of pesticide has arrived in the target region.

[0133] Preferably, the data set also includes information on the applied quantities, i.e. the quantities dispensed by the dosing unit.

[0134] Preferably, the data set also includes information on the times at which drops were dispensed by the dosing unit and on the times at which images, preferably one or more third images, were generated.

[0135] Preferably, the data set also includes information on the geocoordinates of the device according to the invention at the time of identification of the target object and / or at the time of determination / identification of the target region and / or at the time of application of the plant protection agent and / or at another time and / or at several other times before, during, and / or after the execution of the method according to the invention. To determine the geocoordinates, the device according to the invention can be equipped with means for determining the position, such as a GPS sensor (GPS: Global Positioning System) or another sensor for determining the position based on a navigation satellite system.

[0136] The data set can be stored in a data storage device. It can be displayed to a human using a display device (e.g., a monitor or printer). It can be transmitted to a (separate) computer system via a transmitter. The transmission can be entirely or partially wired and / or wireless (e.g., via mobile communications).

[0137] The invention is explained in more detail below with reference to drawings, without wishing to limit the invention to the features and combinations of features shown in the drawings.

[0138] Fig. 1 shows an example and schematically the method according to the invention in the form of four steps in a flow chart.

[0139] The method (100) comprises the steps:

[0140] (110) Identify the target object and determine the application rate

[0141] (120) Determining a target region

[0142] (130) Applying a plant protection product to the target region (140) Documenting the application

[0143] Fig. 2 shows, by way of example and schematically, a preferred embodiment of the method according to the invention in the form of a flow chart.

[0144] The method (200) comprises the steps:

[0145] (210) generating a first image of a target object,

[0146] (220) Identifying the target object based on the first image capture,

[0147] (230) Determining a target region based on the first image recording and / or on the basis of a second image recording,

[0148] (240) Applying a quantity of a plant protection product to the target region using a dosing unit,

[0149] (250) generating a third image, wherein the third image shows the plant protection product applied to the target region,

[0150] (260) Creating a data set comprising

[0151] • the first and / or second image capture,

[0152] • information about the identified target object, and

[0153] • the third image capture,

[0154] (270) Outputting and / or storing the data set and / or transmitting the data set to a computer system .

[0155] Fig. 3 shows an example and schematically a preferred embodiment of the device according to the invention.

[0156] The device (1) comprises a control and computing unit (2), at least one image recording unit (3, 4), a dosing unit (5), and a data memory (6).

[0157] The device shown in Fig. 3 comprises two image recording units, a first image recording unit (3) and a second image recording unit (4). The first image recording unit (3) can be used to generate a first image recording or a plurality of first images that at least partially show a target object, and on the basis of which the target object is identified. The second image recording unit (4) can be used to generate a second and / or third image recording or a plurality of second and / or third image recordings. The one or more second image recordings serve to determine / identify the target region; the one or more third image recordings serve to verify the exclusive application of plant protection agent to the target region. The device (1) can optionally comprise a transmitting unit (7) and / or a position-determining unit (8) and / or a source (9) for electromagnetic radiation.

[0158] The control and computing unit (2) is configured to receive a first image from the at least one first image recording unit (3), to identify a target object in the first image recording, optionally to receive a second image recording from the at least one second image recording unit (4), to determine a target region based on the first and / or second image recording, to cause the dosing unit (5) to apply a quantity of a plant protection agent to the target region, to cause the at least one second image recording unit (4) to generate a third image recording, wherein the third image recording shows the plant protection agent applied to the target region, to generate a data set, wherein the data set comprises o the first and / or the second image recording, o information about the determined target object, and o the second image recording, and to store the data set in the data memory (6) and optionally to cause the transmitting unit (7),to transmit the data set to a computer system.,

[0159] The position determination unit (8) can be used to determine the position of the device (1) according to the invention.

[0160] The source (9) for electromagnetic radiation can be used to illuminate the target object and / or the target region in order to produce images independent of the ambient light (e.g. sunlight).

[0161] The device according to the invention can generally (i.e., not limited to the embodiment shown in Fig. 3) have means by which the device can move autonomously within a field for cultivated plants. For example, the device according to the invention can be designed as a robot or a vehicle. Corresponding means of movement are described in the prior art (see, for example: WO2016 / 116888A1, US20030229425A1, WO2019 / 100118A1,

[0162] The device according to the invention can generally (i.e., not limited to the embodiment shown in Fig. 3) have a passive and / or active wind shield. A wind shield can ensure that individual droplets do not drift from the desired target trajectory on their trajectory from the dosing unit toward the target region. Passive wind shielding can be achieved, for example, with protective shields. Active wind shielding can influence the air flow in the area of ​​the dosing unit, e.g., with pneumatic counterpressure. For this purpose, the local wind speed and gust factor can be measured. A model-based prediction algorithm can control the pneumatic system.

[0163] The present invention combines the precise application of plant protection products with the proof that plant protection products have been applied exclusively to defined target regions.

[0164] By means of the present invention, plant protection agents can be dosed in minute quantities in such a way that contamination of groundwater, soil and / or crops is avoided.

[0165] The present invention allows weeds to be selectively inhibited from growing or to promote their growth based on synergistic effects. This makes it possible to maximize biodiversity while taking overall economic efficiency into account in crop production.

[0166] Pest populations do not have to be completely removed, but can be reduced to an ecologically sensible population size.

[0167] To improve the quality of the crop and / or biodiversity, the method according to the invention can be used at regular intervals, preferably daily, with preferably one or more autonomous robots and / or vehicles to enable continuous plant protection and thus to detect and resolve problems at an early stage.

[0168] This invention makes it possible, in principle, to use chemically synthesized pesticides even in certified organic farming, as potentially contaminated plants can be transparently identified through the documentation of every single drop. Digitalization in plant production is intended to provide end customers with information about the distance from their product (crop), the quantities, the substance used, at what times, and under what environmental conditions.

Claims

The procedure includes the following steps: (a) Producing an initial image of a target object, (b) Identifying the target object based on the first image acquisition, (c) Determining a target region based on the first image acquisition and / or based on a second image acquisition, (d) Applying a quantity of a plant protection product to the target area using a metering unit, (e) Producing a third image, wherein the third image shows the pesticide applied to the target region, (f) Generating a data set comprising a. the first and / or the second image acquisition, b. information about the identified target object, and c. the third image acquisition, (g) Output and / or storage of the data set and / or transmission of the data set to a computer system. Method according to claim 1, wherein in step (c) the target region is determined based on the first and second image acquisitions, which are each acquired from different directions, and wherein in step (f) the data set additionally comprises the second image acquisition and depth information relating to a distance of the dosing unit from the target region, which was obtained from the two image acquisitions. Method according to claim 1 or 2, wherein in step (d) the application of the plant protection product in the form of one or more individual drops is carried out. Method according to claims 1 to 3, wherein in step (e) the third image acquisition shows one or more individual drops that have been applied to the target region in step (d).Method according to claims 1 to 4, wherein the data set generated in step (f) additionally includes the total quantity of the pesticide applied in the target region. Method according to claim 5, wherein, based on the data set generated in step (f) and the third image recorded therein, it is determined whether and what quantity of pesticide has arrived in the target region. Method according to claims 1 to 6, wherein in step (d) the dosing unit has been installed on a windbreak, which ensures that one or more individual drops of the... The device comprises a control and computing unit, at least one image acquisition unit, a dosing unit, and a data storage unit and / or a transmission unit, wherein the control and computing unit is configured to receive a first image from the at least one image acquisition unit, identify a target object in the first image, optionally receive a second image from the at least one image acquisition unit, determine a target region based on the first and / or second image, cause the dosing unit to apply a quantity of a plant protection product to the target region, cause the at least one first image acquisition unit to generate a third image, wherein the third image shows the plant protection product applied to the target region, and generate a data record.wherein the data set comprises the first and / or the second image acquisition, information about the identified target object, and the second image acquisition, and to store the data set in the data storage and / or to cause the transmitting unit to transmit the data set to a computer system. Device according to claim 9, wherein the control and computing unit is configured to receive a second image acquisition from a second image acquisition unit, wherein the second image acquisition unit is spatially positioned differently relative to the first image acquisition unit, to determine a target region based on the first and second image acquisition, to cause the dosing unit to apply a quantity of a pesticide to the target region, to cause at least one first image acquisition unit to produce a third image acquisition, wherein the third image acquisition shows the pesticide applied to the target region, to produce a data set, wherein the data set comprises the first image acquisition, information about the identified target object, and the second image acquisition.

10. Device according to claim 8 or 9, wherein the dosing unit comprises a reservoir, a nozzle, a precision valve and a microreservoir suitable for applying drops with a volume of 1 to 100 nL.

11. Device according to claim 10, wherein the drop has a volume of 1 to 50 nL.

12. Device according to claims 9 to 11, wherein the first image acquisition unit and the second image acquisition unit are configured as a stereo camera.

13. Device according to claims 8 to 12, wherein the metering unit is installed on a windscreen which ensures that individual droplets of the pesticide do not drift from the desired trajectory on their flight path from the metering unit towards the target region.

Images