Physical weeding method and system based on double closed-loop feedback

Abstract

The invention provides a physical weeding method and system based on double closed-loop feedback, and belongs to the technical field of intelligent agriculture. The method comprises the steps that before electroporation operation is executed, a target effectiveness index is calculated through first-time bioelectrical impedance measurement, and accurate contact with weed meristem targets is ensured in a closed-loop feedback mode; after the contact is effective, generating optimal electric pulse parameters in a self-adaptive manner according to physiological characteristics of weeds, and executing operation; after operation, the irreversible cracking index of the cells is calculated through subsequent bioelectrical impedance measurement, the operation effect is quantitatively evaluated, and supplementary strike is triggered. The problems that existing physical weeding is low in precision, high in energy consumption and unreliable in effect are solved through double closed loops of verification before contact and confirmation after effect in combination with an adaptive energy strategy, and the precision, reliability and energy utilization efficiency of operation are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of intelligent agriculture technology, in particular to a physical weeding method and system based on double closed-loop feedback. BACKGROUND

[0002] With the development of robot technology and artificial intelligence technology, automated physical weeding has emerged, aiming to replace chemical agents with physical means such as machinery, heat energy, electric energy or laser. Among them, the target electroporation technology based on machine vision guidance is a new and efficient physical weeding method. This method usually uses a deep learning model to analyze the images captured by the camera, identifies the position and morphology of the weeds, and then guides the end effector such as a probe to move to the target position, destroys the cell structure of the meristem tissue of the weeds by applying high-voltage electric pulses, and thus achieves the purpose of precise removal.

[0003] However, the physical weeding method based on high-precision visual positioning in the prior art still has the following technical bottlenecks to be solved in practical application:

[0004] (1) Lack of contact effectiveness verification mechanism: the existing visual guidance system defaults the effectiveness of physical contact after completing target positioning, i.e. starts to perform weeding operation, but due to environmental and equipment errors, there is often a deviation between visual coordinates and actual physical contact points, resulting in the end effector failing to accurately touch the preset meristem target point as the key striking position, leading to a high failure rate of the operation;

[0005] (2) Lack of adaptability of energy management strategy: to cope with individual differences of weeds and try to improve the success rate of single strike, the existing technology generally adopts a fixed overkill energy strategy, i.e. the energy parameter setting value is high, without considering the specific physiological characteristics of the target, causing significant energy redundancy for most standard weeds, resulting in low overall energy efficiency of the system and limiting its application economy;

[0006] (3) Lack of quantitative confirmation of operation effect: after energy application, it is usually assumed that the operation is successful, and there is no mechanism for objectively evaluating the weeding effect, which cannot determine whether the target tissue has been irreversibly damaged by the strike, and there is a risk of recovery of the damaged weeds, making it difficult to guarantee the final reliability of the operation.

[0007] Therefore, we need to develop a physical weeding method and system based on double closed-loop feedback, which can significantly improve the accuracy, reliability and energy utilization efficiency of physical weeding operation through double closed-loop of pre-contact verification and post-effect confirmation combined with adaptive energy strategy. SUMMARY

[0008] The present application aims to provide a physical weeding method and system based on double closed-loop feedback to solve the problems of low precision, poor reliability and energy consumption redundancy caused by the lack of contact effectiveness verification and effect quantification confirmation in the existing weeding solutions mentioned in the background.

[0009] To achieve the above-mentioned purpose, in one aspect, the present application provides a physical weeding method based on double closed-loop feedback, which is specifically as follows:

[0010] Step S1: Multi-modal sensing is performed on the target area, and the meristem target point of the target is located based on the sensing result, so as to obtain a target data packet containing the three-dimensional coordinates and physiological characteristics of the meristem target point;

[0011] Step S2: A first complex electrical impedance measurement vector representing the physical contact state of the actuator probe with the meristem target point is obtained, and a target effectiveness index for quantifying the inherent similarity between the first complex electrical impedance measurement vector and a pre-stored standard bioimpedance feature vector is calculated based on the first complex electrical impedance measurement vector and the target data packet, so as to verify the effectiveness of the physical contact. If the physical contact is invalid, the position of the actuator probe is adjusted and the step is repeated until the physical contact is effective. The calculation of the target effectiveness index includes:

[0012] a. Calculate a key point feature matching index to evaluate the matching degree of the first complex electrical impedance measurement vector with the standard value at a preset feature frequency;

[0013] b. Calculate a spectral pattern coherence index to evaluate the similarity of the gradient change trend of the first complex electrical impedance measurement vector in the entire measurement spectrum with the standard spectrum;

[0014] c. Calculate a discriminant space projection index to evaluate the distance between the high-dimensional feature vector extracted from the first complex electrical impedance measurement vector and the standard projection coordinates after projection to a preset discriminant subspace;

[0015] d. Weighted fusion of the key point feature matching index, the spectral pattern coherence index and the discriminant space projection index is performed to generate the target effectiveness index;

[0016] Step S3: When the physical contact is effective, a set of optimal electric pulse parameters is generated through an adaptive energy generation algorithm based on the physiological characteristics contained in the target data packet, and an electroporation operation is performed on the meristem target point through the actuator probe based on the electric pulse parameters;

[0017] Step S4: After the electroporation operation, based on the multi-stage complex electrical impedance measurement, a cell irreversible lysis index is calculated to confirm the operation effect of the electroporation operation;

[0018] Step S5: Based on the operation effect, if it is confirmed that the operation is not successful, steps S3 and S4 are repeatedly executed until the operation effect is confirmed to be successful.

[0019] Based on the foregoing scheme, the calculation formula of the key point feature matching index in step S2 is as follows:

[0020]

[0021] wherein, and are the normalized deviations of the modulus and phase angle of the first complex electrical impedance measurement vector from the standard values at the preset feature frequency, respectively; and are weight coefficients, which are adaptively generated based on the physiological characteristics of the target data packet including weed species S, growth stage P and / or tissue water content H through a preset weight decision function W(S, P, H).

[0022] Based on the foregoing scheme, the calculation process of the spectral morphology coherence index in step S2 includes:

[0023] e. Calculate the measurement impedance modulus value vector composed of the impedance modulus values obtained at N different measurement frequency points extracted from the first complex electrical impedance measurement vector and the corresponding standard modulus value vector Gradient vector and ;

[0024] f. Based on the cosine similarity, the directional consistency of the gradient vectors and is measured, and the calculation formula of the spectral morphology coherence index is: wherein, represents vector dot product, represents the L2 norm of the vector;

[0025] g. Map to the interval [0, 1] to obtain the spectral morphology coherence index as follows: .

[0026] Based on the foregoing scheme, the calculation process of the discriminant space projection index in step S2 includes:

[0027] h. High-dimensional feature construction: based on the first complex impedance measurement vector, a high-dimensional feature vector containing the modulus, phase angle, conductivity, and dielectric constant is extracted for the target ;

[0028] i. Discriminant projection transformation: a discriminant projection matrix obtained by offline training is used to perform linear projection transformation on the high-dimensional feature vector and the standard high-dimensional feature vector , to obtain and , where and are the projection coordinates of the high-dimensional feature vector and the standard high-dimensional feature vector in the discriminant subspace ;

[0029] j. Projection distance calculation and index generation: the normalized Euclidean distances of the projection coordinates and are calculated as the discriminant space projection index .

[0030] Based on the foregoing scheme, the effectiveness of the physical contact in step S2 is verified, specifically including:

[0031] The target effectiveness index is compared with a preset multi-level threshold value to obtain a graded effective contact or an ineffective contact; wherein the graded effective contact indicates that the physical contact point this time is the target point of the meristematic tissue, including:

[0032] An excellent matching threshold value and an acceptable matching threshold value are set, wherein the excellent matching threshold value is higher than the acceptable matching threshold value ; if the target effectiveness index , the graded effective contact is an excellent pass; if , the graded effective contact is a qualified pass

[0033] If , the ineffective contact is obtained, indicating that the physical contact point this time is not the target point of the meristematic tissue.

[0034] Based on the foregoing scheme, in step S3, a set of optimal electric pulse parameters is generated by an adaptive energy generation algorithm, which is based on the physiological characteristics of the target category S, growth stage P, and tissue water content H contained in the target data packet, and from a preset energy parameter model library, a set of electric pulse parameters that accurately match the physiological characteristics is calculated and generated;

[0035] The adaptive energy generation algorithm also dynamically adjusts the generated electrical pulse parameters based on the graded effective contact.

[0036] Based on the aforementioned scheme, the dynamic adjustment specifically includes:

[0037] If the graded effective contact is excellent, it indicates that the physical contact is precise and effective. Then, a first set of energy parameters that precisely matches the target physiological characteristics is generated. This set of parameters corresponds to the minimum effective energy threshold for achieving irreversible electroporation.

[0038] If the graded effective contact is qualified, it indicates that the physical contact is effective but there may be deviation. Then, a second energy parameter set with a preset safety margin is generated. The energy setting value of the second energy parameter set is higher than that of the first energy parameter set, so as to apply energy while prioritizing the success rate of electroporation operation.

[0039] Based on the aforementioned scheme, the multi-stage complex impedance measurement in step S4 includes:

[0040] A very short first preset time point after the completion of the electroporation operation The actuator probe is instructed to perform a second complex impedance measurement to obtain a second complex impedance measurement vector, aiming to capture the transient physical perforation effect of electroporation on cell tissue at the meristematic target site; at the first preset time point Afterwards, at the second preset time point after a preset relaxation time window. The actuator probe is instructed to perform a third complex impedance measurement to obtain the third complex impedance measurement vector.

[0041] Based on the aforementioned scheme, the irreversible cell lysis index mentioned in step S4 is calculated using a dynamic lysis assessment model based on the first, second, and third complex electrical impedance measurement vectors. The calculation formula is as follows:

[0042]

[0043] in, This refers to the first complex impedance measurement vector obtained before the electroporation operation in step S2; This refers to the first preset time point after the electroporation operation is completed in step S4. The obtained second complex impedance measurement vector is designed to capture the transient physical perforation effect of cells; In step S4, after a preset relaxation time window, at the second preset time point The obtained third complex impedance measurement vector is intended to assess the state stability of the cell; The second preset time point With the first preset time point The time difference between them; These are the weighting coefficients; As a stability factor; For ambient temperature and humidity; Standard temperature and humidity; This is an environmental correction factor; This is a target type correction factor; this formula comprehensively evaluates the magnitude of the decrease in electrical impedance at the target meristem target site, the stability of the lysis process, and compensates for environmental and species differences.

[0044] On the other hand, the present invention also provides a physical weeding system based on dual closed-loop feedback, comprising:

[0045] The perception and localization module is used to implement step S1 and is responsible for generating the target data packet.

[0046] Verification and Decision Module: This module is used to implement steps S2 and S4. It contains a bioelectrical impedance sensor and a core algorithm submodule responsible for calculating the target effectiveness index and the cell irreversible lysis index.

[0047] Energy execution module: used to implement step S3, including an adaptive energy generation algorithm unit and a high-voltage pulse generator;

[0048] Task Control and Navigation Module: Used to implement step S5 and coordinate the work of various modules.

[0049] The present invention has the following advantages and effects compared with the prior art:

[0050] (1) Improved the accuracy and success rate of the operation: By establishing a contact effectiveness verification before electroporation operation, the physical contact is quantitatively judged in real time based on the target effectiveness index, which overcomes the blindness of execution caused by the positioning deviation of the existing technology and ensures accurate contact with the preset target point.

[0051] (2) Reduced system operating energy consumption: Through the adaptive energy generation algorithm, the optimal energy parameters can be generated based on the physiological characteristic data of the target, replacing the fixed high threshold energy strategy of the existing technology, realizing accurate energy matching and significantly improving energy utilization efficiency;

[0052] (3) Ensures the reliability and thoroughness of the operation: After confirming the effect of electroporation, the operation effect is quantitatively evaluated based on the irreversible cell lysis index. The irreversible damage to the target can be confirmed and a supplementary attack can be triggered, thus ensuring the final effect of weed removal. Attached Figure Description

[0053] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0054] Figure 1 This is a flowchart of a physical weeding method based on dual closed-loop feedback provided in an embodiment of the present invention;

[0055] Figure 2 This is a structural diagram of a physical weeding system based on dual closed-loop feedback provided in an embodiment of the present invention. Detailed Implementation

[0056] To more clearly explain the purpose, technical solutions, and advantages of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein. On the contrary, these embodiments are provided so that the present invention will be more comprehensive and complete, and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0057] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a full understanding of embodiments of the invention. However, those skilled in the art will recognize that the technical solutions of the invention can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of the invention.

[0058] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0059] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0060] The present invention will now be described in detail with reference to specific embodiments:

[0061] Example 1

[0062] As attached Figure 1 As shown, Embodiment 1 of the present invention provides a physical weeding method based on dual closed-loop feedback. The specific steps of the method are as follows:

[0063] Step S1: Perform multimodal perception on the target area to obtain a target data package containing three-dimensional coordinates and physiological features;

[0064] Specifically, the multimodal perception is achieved by using a drone platform equipped with a high-definition RGB camera and a multispectral camera to perform a cruise scan along a preset path over a designated work area, i.e., a target area, to collect and form a raw image data stream in real time.

[0065] Furthermore, the original image data stream is input into a weed recognition model pre-trained by deep learning. This model is based on a convolutional neural network architecture and can analyze the data stream to accurately identify the species (e.g., Chenopodium album), physical size (e.g., crown width, plant height), and growth stage (e.g., 4-6 leaf stage) of individual weeds.

[0066] Furthermore, the identification results and the corresponding raw image data stream are input into a plant anatomy localization model. The core function of this model is to locate the physiologically most vulnerable meristematic tissue target point of the weed, suitable for subsequent attack, based on prior knowledge of the weed species and growth stage. It then calculates the three-dimensional spatial coordinates of this meristematic tissue target point by combining real-time differential GPS pose data from the UAV platform and camera intrinsic and extrinsic parameters. Simultaneously, the water content of the target tissue is estimated based on data collected by the multispectral camera of the UAV platform.

[0067] It should be noted that the meristematic tissue of a plant is the center of its growth and development, containing a large number of continuously dividing cells. By using the meristematic tissue as a target, irreversible damage to weeds can be achieved with minimal energy expenditure, thereby achieving complete eradication. This is the key premise for the precise and efficient operation of this invention.

[0068] Specifically, the localization logic for the meristematic target points in the plant anatomical localization model is executed according to the following differentiated strategy:

[0069] For monostem or broadleaf weeds: if they are in the early stage of vegetative growth (such as before the 4-6 leaf stage), their main growth point, i.e., the apical meristem of the stem tip, is taken as the target meristem; if they are in the middle and late stages of vegetative growth (i.e., the 4-6 leaf stage and later), the apical meristem is preferentially located as the target meristem. However, if it cannot be reliably identified due to severe shading or physical damage, the axillary bud meristem located in the uppermost leaf axil and which has already developed is identified as the target meristem.

[0070] For clump-forming or grassy weeds: If they are in the tillering stage or earlier, since all growth points are concentrated at the base, the basal meristem (tillering nodes) near the ground surface of the plant is used as the target meristem; if they are in the jointing stage or later, the intercalary meristem closest to the ground and already activated is used as the target meristem to effectively prevent the elongation of the stem.

[0071] After this step, for each identified and successfully located target (in this embodiment, weeds), a structured target data packet is finally output. The target data packet is an independent dataset, including the following data fields: a unique identifier for the meristematic target; the three-dimensional spatial coordinates of the meristematic target; the target type S (in this embodiment, weed type); the growth stage P of the target (in this embodiment, the weed growth stage); and the estimated tissue water content H of the meristematic target (in this embodiment, the weed tissue water content).

[0072] Step S2: Obtain the first complex impedance measurement vector characterizing the physical contact state between the actuator probe and the target. Based on the first complex impedance measurement vector and the target data packet, calculate the target effectiveness index. Use this index to determine whether the current physical contact point is the meristematic tissue target. If so, the current physical contact is effective. If not, start closed-loop feedback control, adjust the position of the actuator probe, and repeat step S2 until the contact is effective.

[0073] It should be noted that step S2 is the core innovation of this invention. It is a pre-contact verification loop before applying any electroporation energy to the meristematic target of the weed. This is used to absolutely and objectively evaluate whether the end of the actuator probe has made physical contact with the predetermined and effective meristematic target.

[0074] Specifically, the process of obtaining the first complex impedance measurement vector includes: based on the three-dimensional spatial coordinates of the meristematic target point in the target data, the visual servo control unit precisely controls an actuator probe with an integrated bioelectrical impedance sensor at its end to move towards the target. At the instant the actuator probe tip makes physical contact with the target tissue surface, the bioelectrical impedance sensor is immediately triggered, performing the first high-frequency multi-frequency scanning measurement operation (e.g., the measurement frequency range is from 10kHz to 1MHz, selecting N=10 logarithmically spaced frequency points, and the measurement spectrum is from...). arrive This yields a real-time, original, first complex impedance measurement vector.

[0075] Preferably, the target effectiveness index is used to quantify the intrinsic similarity between the first complex impedance measurement vector and the pre-stored standard bioimpedance characteristic vector, and its calculation steps include:

[0076] Step S201: Adaptive weighted calculation of key point feature matching index This method is used to evaluate the degree of matching between the corresponding component in the first complex impedance measurement vector and the standard value at the characteristic frequency (e.g., 500 kHz) that is most effective in distinguishing tissue types. The characteristic frequency is obtained through previous experimental calibration, including: performing multi-band bioelectrical impedance scanning on various known weed samples (such as meristematic tissue, mature leaves, etc.), finding the most significant frequency point that distinguishes different tissue types through data analysis (such as linear discriminant analysis LDA), and determining it as the characteristic frequency.

[0077] Furthermore, by utilizing an adaptive weighting method, the key point feature matching index is... This allows for targeted diagnoses for different objectives, significantly improving the accuracy of assessments. Keypoint Feature Matching Index The calculation formula is as follows:

[0078]

[0079] in, and These are the normalized deviations of the magnitude and phase angle of the first complex impedance measurement vector from the standard value at a preset characteristic frequency, respectively. and These are the weighting coefficients;

[0080] Specifically, the modulus deviation and phase deviation The calculation formula is as follows: , ,in , These are the impedance magnitude and phase angle obtained from the first complex impedance measurement at the characteristic frequency, respectively. , This refers to the standard impedance magnitude and phase angle corresponding to this characteristic frequency, retrieved from the database.

[0081] Preferably, the weighting coefficient and The method for determining this is the innovation of step S201, specifically through a weighted decision function. The function is generated in real time, and its inputs are the target species S, growth stage P, and tissue water content H in the target data packet.

[0082] Specifically, the target species S is processed into a uniquely thermally encoded vector, the growth stage P is normalized to a value between 0 and 1 (e.g., 0.4 for the 4-leaf stage), and the tissue water content H is an estimated percentage value.

[0083] Specifically, the weighted decision function The underlying logic is that for weed species with thicker cell walls and higher lignification, or targets in their mature stage, the impedance modulus better reflects their physical structure. The weight will be adjusted accordingly; for young meristematic tissues with active cells and high water content, the phase angle is extremely sensitive to the capacitance effect of the cell membrane, therefore The weight will be significantly increased.

[0084] For example, to achieve adaptive weight allocation, the weight decision function is specifically implemented as a pre-trained regression model (e.g., a gradient boosting decision tree), which is trained to output a single parameter. And based on this, calculate the weighting coefficients as follows: Let , ;

[0085] In a specific embodiment of the present invention, when the target organism is a weed, the above logic is specifically manifested as follows: when the target is a mature weed (e.g., a high degree of lignification corresponding to P and S), the regression model will output a larger value. Values ​​(such as 0.8), thus making , The model focuses on modulus evaluation, which is more sensitive to physical structure; when the target is immature meristem (e.g., small P-value, high H-value), the regression model will output a smaller value. Values ​​(such as 0.3), thus making , It focuses on phase angle assessment, which is more sensitive to cell activity.

[0086] The above design transforms the complex weight allocation problem into a single-value regression task with clear physical meaning, greatly improving the feasibility and reliability of the adaptive mechanism.

[0087] Step S202: Calculate the spectral morphology coherence index The entire measurement spectrum (from) is used to evaluate the first complex impedance measurement vector. arrive The overall shape of the spectrum is similar to that of the standard spectrum.

[0088] It should be noted that this step transcends the limitations of single-point comparison. Its innovation lies in comparing not the impedance values ​​directly, but rather their trend in the frequency domain, i.e., the derivative of the spectrum. This includes the following:

[0089] Data preprocessing: Extract the measured impedance magnitude vector from the first complex impedance measurement vector, which consists of the impedance magnitudes obtained at N different measurement frequency points. and the corresponding standard modulus vector The gradient vectors are obtained by performing first-order differences on the logarithmic frequency axis. and .

[0090] Coherence calculation: Cosine similarity is used to measure the gradient vector. and The method ensures directional consistency; it is insensitive to the absolute magnitude of the vector, satisfies the requirement of comparing shape rather than amplitude, and obtains a preliminary spectral morphology coherence index. The calculation is as follows:

[0091]

[0092] in, Represents the dot product of vectors. The L2 norm of a vector; to unify the dimensions, Mapping to the [0,1] interval yields the spectral morphology coherence index. as follows: .

[0093] Step S203: Calculate the discriminant spatial projection index This method is used to project tissue types (such as meristematic tissue, mature leaves, and lignified stems of weeds) that may be difficult to distinguish in the original high-dimensional feature space onto a new, low-dimensional discriminative subspace through discriminative projection transformation, thereby maximizing the inter-class distance between different tissue categories in this subspace. Specifically:

[0094] High-dimensional feature construction: Based on the first complex impedance measurement vector, a high-dimensional feature vector containing parameters such as magnitude, phase angle, conductivity, and dielectric constant is extracted for the target, i.e., weeds. Similarly, based on the standard bioimpedance eigenvector, a standard high-dimensional eigenvector is constructed for standard meristems. .

[0095] Discriminant projection transformation: using a discriminant projection matrix obtained through pre-trained offline training. The high-dimensional feature vector constructed in the preceding steps and standard high-dimensional feature vectors Perform a linear projection transformation to obtain and ,in and These are high-dimensional feature vectors. and standard high-dimensional feature vectors The projection coordinates in the discriminant subspace. The method for obtaining the discriminant projection matrix W includes: collecting a large number of weed samples with known tissue categories (such as meristematic tissue, mature leaves, lignified stems, etc.), measuring their complex electrical impedance and extracting high-dimensional feature vectors to form a labeled training dataset; then, using machine learning algorithms such as linear discriminant analysis (LDA) to train the dataset to obtain an optimal projection matrix that maximizes the inter-class distance and minimizes the intra-class distance after projection of samples of different categories, which is used as the discriminant projection matrix W.

[0096] Projection distance calculation and exponent generation: Calculating projected coordinates and The normalized Euclidean distance is used as the discriminant spatial projection index. Its evaluation power far exceeds any direct comparison in the original feature space, and the calculation formula is as follows:

[0097]

[0098] Step S204: Integrate and generate the final target effectiveness index. The three sub-indices with clear physical and mathematical meanings are the key point feature matching indices. The spectral morphology coherence index The discriminant spatial projection index The target effectiveness index is obtained by performing weighted fusion. The calculation formula is as follows: Among them, the fusion weight The sum of 1 is derived through grid search on a validation set containing thousands of labeled valid or invalid physical contact samples, with the optimization objective being to maximize the F1 score of the validation classification. In this embodiment, a set of optimized weights is: , , This reflects that in the judgment, the confidence in the accuracy of key points is the highest, followed by overall morphological similarity and depth discrimination features.

[0099] Step S205, Multi-level threshold determination and closed-loop control: The target effectiveness index... The effective or ineffective contact is determined by comparing the results with the preset multi-level thresholds.

[0100] Specifically, if the graded effective contact is obtained, it indicates that the physical contact point is the meristematic target point, including: setting an excellent matching threshold. Acceptable matching threshold ,like If effective contact is graded, it is considered excellent; if If the graded effective contact is then considered qualified and passes;

[0101] Specifically, if If the contact is invalid, it means that the physical contact point is not the target point of the meristematic tissue. Closed-loop feedback control is then activated, instructing the actuator probe to make micron-level physical position adjustments (e.g., moving 50 microns along a preset spiral search path). After the position adjustment is completed, step S2 is restarted until the judgment result is excellent or qualified before proceeding to the next step S3.

[0102] Step S3: After the physical contact is effective, based on the physiological characteristics contained in the target data packet, an optimal set of electrical pulse parameters is generated through an adaptive energy generation algorithm. Based on the electrical pulse parameters, an electroporation operation is performed on the target through an actuator probe to maximize energy utilization efficiency.

[0103] This step, as the execution link of energy application, has the overall goal of dynamically and intelligently calculating and generating a set of optimized electroporation energy parameters, i.e., electrical pulse parameters, based on the specific physiological characteristics of the target, after confirming in the previous step that the contact with the meristematic tissue target is absolutely effective, in order to maximize energy utilization efficiency.

[0104] Specifically, the input of the adaptive energy generation algorithm is the target data packet output in step S1. Based on the physiological characteristics such as target type S, growth stage P and tissue water content H contained in the target data packet, the algorithm calculates and generates a set of optimal electrical pulse parameters that precisely match the physiological characteristics from a preset energy parameter model library.

[0105] Preferably, the adaptive energy generation algorithm further dynamically adjusts the generated electrical pulse parameters based on the graded effective contact output in step S2. Specifically:

[0106] If the graded effective contact is rated as excellent, indicating that the physical contact is precise and effective, a first set of energy parameters that precisely matches the target physiological characteristics is generated. This set of parameters corresponds to the minimum effective energy threshold for achieving irreversible electroporation.

[0107] If the graded effective contact is qualified, it indicates that the physical contact is effective but there may be deviations. Then, a second energy parameter set with a preset safety margin is generated. The energy setting value of the second energy parameter set is higher than that of the first energy parameter set, so as to apply energy while prioritizing the success rate of electroporation operation.

[0108] For example, the energy parameter model library is a multiple linear regression model. ,in For voltage, The pulse width. The core function of this model, based on the physiological characteristics of the target obtained through the target data packet, is to calculate the minimum effective energy threshold for irreversible electroporation, i.e., the first energy parameter set. The model is constructed by: through extensive preliminary experiments, calibrating the minimum effective electrical pulse parameter combination that can cause irreversible damage to the meristematic tissue of target samples (i.e., weed samples) with different target species S, growth stages P, and tissue water content H. and the large number of data points collected As input features, The energy parameter model library is obtained by using the output labels to train one or more regression models (such as multiple linear regression, gradient boosting trees, etc.).

[0109] In practice, the model works in conjunction with the graded effective contact: it calls a multiple linear regression model to calculate the first set of energy parameters. For example, if the target is quinoa (S-coded as 1), 4-leaf stage (P=4), and water content 85% (H=85), the model output is the first set of energy parameters: {voltage amplitude 3000V, pulse width 100ns, pulse count 50}, thus achieving precise energy matching. Then, a decision is made based on the graded effective contact: if the graded effective contact is excellent, the electroporation operation is directly performed using the above first set of energy parameters; if the graded effective contact is acceptable, a preset safety margin is applied to the first set of energy parameters to generate a second set of energy parameters. For example, the voltage amplitude is increased by 20% to 3600V, resulting in the final second set of energy parameters {3600V, 100ns, 50 pulses}. In this way, both precise energy matching based on physiological characteristics and dynamic risk compensation based on physical contact quality are achieved.

[0110] Furthermore, the electroporation operation is based on the electrical pulse parameters calculated by the adaptive energy generation algorithm unit, and uses a nanosecond-level high-voltage pulse generator to perform an irreversible electroporation operation on the meristematic target point through the actuator probe.

[0111] Step S4: Based on multi-stage complex impedance measurement, calculate the irreversible cell lysis index at the meristematic target site using a dynamic lysis evaluation model to determine whether the electroporation operation has reached the standard of irreversible biological damage.

[0112] This step is one of the core innovations of this invention, serving as a closed-loop confirmation of the effect based on dynamic physical process modeling. Specifically, it is as follows:

[0113] Preferably, the multi-stage complex impedance measurement includes:

[0114] A very short first preset time point after the completion of the electroporation operation The actuator probe is instructed to perform a second complex impedance measurement to obtain a second complex impedance measurement vector, aiming to capture the transient physical perforation effect of electroporation on cell tissue at the meristematic target site; at the first preset time point Afterwards, at the second preset time point after a preset relaxation time window. The actuator probe is instructed to perform a third complex impedance measurement to obtain the third complex impedance measurement vector. This is intended to observe and evaluate the short-term stability of cell tissue after electroporation to determine whether the damage is irreversible. For example, ms, ms.

[0115] Preferably, the dynamic lysis assessment model is used to calculate a comprehensive index, namely the cell irreversible lysis index, based on the first, second, and third complex electrical impedance measurement vectors. Its value is expressed as... express.

[0116] Specifically, the cell irreversible lysis index Calculated using the following formula:

[0117]

[0118] in, It is the first complex impedance measurement vector obtained in step S2 before the electroporation operation; This refers to the first preset time point after the electroporation operation is completed in step S4. The obtained second complex impedance measurement vector is intended to capture the transient physical perforation effect of cells; Z3 is the value obtained in step S4 after a preset relaxation time window at the second preset time point. The obtained third complex impedance measurement vector is intended to assess the state stability of the cell; The second preset time point With the first preset time point The time difference between them; The weighting coefficients are dimensionless and In this embodiment, This indicates that they value it more. The initial impedance drop at time; This is a stability factor, dimensionless, used to amplify or reduce the impact of the cracking process on stability. It is determined experimentally, and here it is taken as... ; For ambient temperature and humidity; Standard temperature and humidity (e.g., 25°C, 60%RH); This is an environmental correction factor, calibrated experimentally, for example... ; The target type correction coefficient is obtained through a lookup table based on empirical data. For example, for a target that is weeds and a category that is lambsquarters, then... For other more tolerant weeds This formula comprehensively assesses the magnitude of the decrease in electrical impedance at the target meristem site, the stability of the cleavage process, and compensates for environmental and species differences.

[0119] Furthermore, the cell irreversible lysis index is... With a preset irreversible cleavage confirmation threshold A comparison is made to obtain a conclusion confirming the effectiveness of the work, specifically including: if If the condition is met, the electroporation operation is considered successful as it has reached the standard for irreversible biological damage; otherwise, it is considered unsuccessful as it has not reached the standard for irreversible biological damage. For example, .

[0120] Step S5: Based on the conclusions of the operation results, conduct supplementary strikes and target switching.

[0121] This step is used to systematically manage the work queue after confirming that the current target, i.e., the weeds, has been completely cleared, and to drive the drone platform to autonomously navigate to the next target to be processed, so as to ensure that all designated tasks in the entire area can be executed in an orderly and complete manner.

[0122] Specifically, the supplementary strike includes: if the operation is unsuccessful, triggering a supplementary strike operation, returning and repeating steps S3 and S4 until the operation is successful.

[0123] Specifically, the target switching means that the removal process for the currently processed target (i.e., a single weed) is only considered truly complete when the operation is successful. At this point, the system removes the target from the global task list or marks it as completed, and autonomously navigates to the next target to be processed in the task list, starting a new round of removal operations from step S2. Once all targets in the task list have been processed, the system determines that all tasks are complete and executes a return command.

[0124] To fully verify the technical effects of the present invention, this embodiment uses a control experiment for illustration.

[0125] 1. Experimental setup:

[0126] Experimental subjects: common broadleaf weeds in farmland, specifically lambsquarters, at the 4-6 leaf stage;

[0127] The present invention embodiment group adopts the method described in embodiment 1.

[0128] Comparative Example Group: This method employs the physical weeding method described in the background section, which is based on visual positioning, uses a fixed overkill energy, and lacks feedback verification of the killing effect. This method also includes weed identification (obtaining target data packets through multimodal perception) and visual guidance of the actuator probe for physical contact in step S1 of Example 1. However, it lacks the contact effectiveness verification before electroporation and the effect confirmation after electroporation as described in Example 1. Furthermore, its striking method involves impacting the target with a high-speed ejected metal probe, with a fixed striking energy of 0.5 joules. This value is an overkill value set to ensure that more than 95% of the strongest samples are killed, lacking adaptive adjustment capability. Furthermore, the operation is assumed to be successful after the strike is completed, without comparative confirmation, and it is a one-way execution, moving to the next target after processing one.

[0129] 2. Experimental Results and Analysis:

[0130] The results of repeated experiments on 100 *Chenopodium album* plants are as follows:

[0131] In the embodiment group of this invention, the weed removal success rate is 100%. The effectiveness verification before electroporation avoids ineffective strikes, and the effect confirmation after electroporation ensures thorough strikes. At the same time, the adaptive energy generation algorithm reduces the average energy consumption by 40% compared with the comparative group.

[0132] The control group had a weed removal success rate of only 68%, with the failure mainly due to initial contact point deviation and the use of excessive killing energy leading to significant energy waste.

[0133] In summary, the experimental results fully demonstrate that the present invention significantly improves the accuracy, reliability, and energy utilization efficiency of physical weeding by constructing a dual closed-loop feedback mechanism that verifies the effectiveness of contact before electroporation and confirms the effect after electroporation, especially the refined modeling and calculation of its core target effectiveness index.

[0134] Example 2

[0135] As attached Figure 2 As shown, Embodiment 2 of the present invention provides a physical weeding system 200 based on dual closed-loop feedback, comprising:

[0136] Perception and localization module 201: Used to implement step S1, responsible for generating target data packets.

[0137] Verification and Decision Module 202: The core of the system, used to implement steps S2 and S4. It internally contains a bioelectrical impedance sensor and a module responsible for calculating the target effectiveness index. The core algorithm submodule for the cell irreversible lysis index D.

[0138] Energy execution module 203: used to implement step S3, including an adaptive energy generation algorithm unit and a high-voltage pulse generator.

[0139] Task Control and Navigation Module 204: The system's overall commander, responsible for implementing step S5 and coordinating the work of each module.

[0140] In this embodiment 2, the functions of each module of the system correspond to the method steps in embodiment 1. Its internal logic and algorithm, especially the complex calculation process of the target effectiveness index and the dynamic rupture evaluation model implemented in the verification and decision module 202, are consistent with those in embodiment 1, and will not be repeated here.

[0141] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims. It should be understood that the invention is not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A physical weeding method based on dual closed-loop feedback, characterized in that, include: Step S1: Perform multimodal perception on the target area, locate the meristematic target point based on the perception results, and thus obtain a target data package containing the three-dimensional coordinates and physiological characteristics of the meristematic target point; Step S2: Obtain the first complex impedance measurement vector characterizing the physical contact state between the actuator probe and the meristematic target point. Based on the first complex impedance measurement vector and the target data packet, calculate a target effectiveness index to quantify the intrinsic similarity between the first complex impedance measurement vector and the pre-stored standard bioimpedance feature vector to verify the effectiveness of the physical contact. If the physical contact is invalid, adjust the position of the actuator probe and repeat this step until the physical contact is effective. The calculation of the target effectiveness index includes: a. Calculate the feature matching index of a key point. The degree of matching between the first complex impedance measurement vector and the standard value at a preset characteristic frequency is evaluated. b. Calculate the coherence index of a spectral morphology. The similarity between the gradient change trend of the first complex impedance measurement vector over the entire measurement spectrum and the standard spectrum is evaluated. c. Calculate the discriminant spatial projection index The distance between the high-dimensional feature vector extracted from the first complex impedance measurement vector and the standard projection coordinates after being projected onto a preset discriminant subspace is evaluated. d. The key point feature matching index, the spectral morphology coherence index, and the discriminative spatial projection index are weighted and fused to generate the target effectiveness index; Step S3: After the physical contact is effective, based on the physiological characteristics contained in the target data packet, an optimal set of electrical pulse parameters is generated through an adaptive energy generation algorithm, and based on the electrical pulse parameters, an electroporation operation is performed on the meristematic tissue target point through the actuator probe; Step S4: After the electroporation operation, the irreversible cell lysis index is calculated based on multi-stage complex impedance measurement to confirm the effectiveness of the electroporation operation. Step S5: Based on the job results, if the job is confirmed to be unsuccessful, repeat steps S3 and S4 until the job results are confirmed to be successful.

2. The physical weeding method based on dual closed-loop feedback according to claim 1, characterized in that, The formula for calculating the key point feature matching index in step S2 is as follows: ,in, and These are the normalized deviations of the magnitude and phase angle of the first complex impedance measurement vector from the standard value at a preset characteristic frequency, respectively. and The weighting coefficients are adaptively generated based on the physiological characteristics of the target data packet, including weed species S, growth stage P, and / or tissue water content H, through a preset weighting decision function W(S, P, H).

3. The physical weeding method based on dual closed-loop feedback according to claim 1, characterized in that, The calculation process of the spectral morphology coherence index in step S2 includes: e. Calculate the measured impedance magnitude vector extracted from the first complex impedance measurement vector, which consists of the impedance magnitudes obtained at N different measurement frequency points. and the corresponding standard modulus vector gradient vector and ; f. Measure the gradient vector based on cosine similarity. and Based on the directional consistency, the formula for calculating the spectral morphology coherence index is: ,in, Represents the dot product of vectors. The L2 norm of a vector; g. will Mapping to the [0, 1] interval yields the spectral morphology coherence index. as follows: .

4. The physical weeding method based on dual closed-loop feedback according to claim 1, characterized in that, The determination of spatial projection index in step S2 The calculation process includes: h. High-dimensional feature construction: Based on the first complex impedance measurement vector, a high-dimensional feature vector containing the magnitude, phase angle, conductivity, and dielectric constant is extracted for the target. ; i. Discriminant projection transformation: using a discriminant projection matrix obtained through pre-trained offline training. For the high-dimensional feature vector and standard high-dimensional feature vectors Perform a linear projection transformation to obtain and ,in and These are high-dimensional feature vectors. and standard high-dimensional feature vectors Projected coordinates in the discriminant subspace; j. Projection distance calculation and exponent generation: Calculating projected coordinates and The normalized Euclidean distance is used as the discriminant spatial projection index. .

5. The physical weeding method based on dual closed-loop feedback according to claim 1, characterized in that, Step S2, which verifies the effectiveness of physical contact, specifically includes: The target effectiveness index Compared with preset multi-level thresholds, graded effective contact or ineffective contact is obtained; wherein the graded effective contact indicates that the physical contact point is the meristematic tissue target point, including: Set a good matching threshold Acceptable matching threshold Among them, the excellent matching threshold Above the acceptable matching threshold If the target effectiveness index If effective contact is graded, it is considered excellent; if If the graded effective contact is then considered qualified and passes; like If the contact is invalid, it means that the physical contact point is not the target point of the meristematic tissue.

6. The physical weeding method based on dual closed-loop feedback according to claim 5, characterized in that, The step S3, which involves generating a set of optimal electrical pulse parameters using an adaptive energy generation algorithm, is based on the physiological characteristics of the target data packet, including the target species S, growth stage P, and tissue water content H. A set of electrical pulse parameters that precisely match the physiological characteristics is calculated and generated from a preset energy parameter model library. The adaptive energy generation algorithm also dynamically adjusts the generated electrical pulse parameters based on the graded effective contact.

7. The physical weeding method based on dual closed-loop feedback according to claim 6, characterized in that, The dynamic adjustment specifically includes: If the graded effective contact is excellent, it indicates that the physical contact is precise and effective. Then, a first set of energy parameters that precisely matches the target physiological characteristics is generated. This set of parameters corresponds to the minimum effective energy threshold for achieving irreversible electroporation. If the graded effective contact is qualified, it indicates that the physical contact is effective but there may be deviation. Then, a second energy parameter set with a preset safety margin is generated. The energy setting value of the second energy parameter set is higher than that of the first energy parameter set, so as to apply energy while prioritizing the success rate of electroporation operation.

8. The physical weeding method based on dual closed-loop feedback according to claim 1, characterized in that, The multi-stage complex impedance measurement in step S4 includes: A very short first preset time point after the completion of the electroporation operation The actuator probe is instructed to perform a second complex impedance measurement to obtain a second complex impedance measurement vector, aiming to capture the transient physical perforation effect of electroporation on cell tissue at the meristematic target site; at the first preset time point Afterwards, at the second preset time point after a preset relaxation time window. The actuator probe is instructed to perform a third complex impedance measurement to obtain the third complex impedance measurement vector.

9. A physical weeding method based on dual closed-loop feedback according to claim 8, characterized in that, The irreversible cell lysis index mentioned in step S4 is calculated using a dynamic lysis assessment model based on the first, second, and third complex electrical impedance measurement vectors. The calculation formula is as follows: ， in, This refers to the first complex impedance measurement vector obtained before the electroporation operation in step S2; This refers to the first preset time point after the electroporation operation is completed in step S4. The obtained second complex impedance measurement vector is designed to capture the transient physical perforation effect of cells; In step S4, after a preset relaxation time window, at the second preset time point The obtained third complex impedance measurement vector is intended to assess the state stability of the cell; The second preset time point With the first preset time point The time difference between them; These are the weighting coefficients; As a stability factor; For ambient temperature and humidity; Standard temperature and humidity; This is an environmental correction factor; This is a target type correction factor; this formula comprehensively evaluates the magnitude of the decrease in electrical impedance at the target meristem target site, the stability of the lysis process, and compensates for environmental and species differences.

10. A physical weeding system based on dual closed-loop feedback, used to perform a physical weeding method based on dual closed-loop feedback as described in any one of claims 1-9, characterized in that, include: The perception and localization module is used to implement step S1 and is responsible for generating the target data packet. Verification and Decision Module: This module is used to implement steps S2 and S4. It contains a bioelectrical impedance sensor and a core algorithm submodule responsible for calculating the target effectiveness index and the cell irreversible lysis index. Energy execution module: used to implement step S3, including an adaptive energy generation algorithm unit and a high-voltage pulse generator; Task Control and Navigation Module: Used to implement step S5 and coordinate the work of various modules.

Images