Method and system for dynamic control of energy output of high-frequency electric shock weeding device

Abstract

The application provides a high-frequency electric shock weeding device energy output dynamic control method and system, relates to the intelligent control technical field, and the method comprises the following steps: according to the temperature rise rate distribution in the thermodynamic response signal, calibrating multiple high-heat response focal points in the energy action area; connecting the high-heat response focal points to construct a dynamic action range boundary, and adjusting the boundary form according to the real-time heat diffusion trend; dividing the action range boundary into several independent energy regulation zones according to the thermodynamic gradient characteristics; generating a partition energy compensation coefficient based on the heat damage degree and energy accumulation state of each regulation zone; differentially adjusting the high-frequency current output according to the partition energy compensation coefficient difference until the thermodynamic response signal meets the global inactivation condition. The application can dynamically adjust the energy output parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, in particular to an energy output dynamic control method and system of a high-frequency electric shock weeding device. BACKGROUND

[0002] The existing electric shock weeding technology has some technical defects. For example, the weeding robot in the test field lacks dynamic control of energy output when applying high-frequency current to weeds. In actual operation, the above-ground morphological characteristics of different weeds are different, the plant size and spatial distribution have great differences, and the thermodynamic response of weed tissues to current is also different.

[0003] Some current technologies use uniform energy parameters for electric shock, which cannot be adjusted according to the specific conditions of weeds. For example, for some weeds with small plant size and high sensitivity to current, too high energy output not only causes energy waste, but also may have unnecessary impact on the surrounding soil environment and crops; and for some weeds with large plant size and developed root system, if the energy output is insufficient, it cannot achieve the effect of complete inactivation, which leads to easy regrowth of weeds. SUMMARY

[0004] The technical problem to be solved by the present application is to provide an energy output dynamic control method and system of a high-frequency electric shock weeding device, which can dynamically adjust energy output parameters.

[0005] To solve the above technical problems, the technical solution of the present application is as follows:

[0006] In a first aspect, an energy output dynamic control method of a high-frequency electric shock weeding device is provided, the method comprising:

[0007] Step 1: identifying the above-ground morphological characteristics of target weeds through a visual sensor to generate positioning data containing plant size and spatial distribution;

[0008] Step 2: calculating the voltage amplitude, frequency and action time parameters of the initial high-frequency current according to the preset mapping relationship based on the plant size in the positioning data;

[0009] Step 3: executing the initial energy parameters and applying high-frequency current to weeds, while collecting the thermodynamic response signals of weed tissues in real time;

[0010] Step 4: calibrating multiple high-heat response foci in the energy action area according to the temperature rise rate distribution in the thermodynamic response signals;

[0011] Step 5: connecting the high-heat response foci to construct a dynamic action range boundary, and adjusting the boundary shape according to the real-time heat diffusion trend;

[0012] Step 6, dividing the action range boundary into several independent energy regulation zones according to the thermodynamic gradient characteristics;

[0013] Step 7, generating a partition energy compensation coefficient based on the thermal damage degree and energy accumulation state of each regulation zone;

[0014] Step 8, differentially adjusting the high-frequency current output according to the partition energy compensation coefficient until the thermodynamic response signal meets the global inactivation condition.

[0015] In a second aspect, the energy output dynamic control system of the high-frequency electric shock weeding device comprises:

[0016] The acquisition module is configured to identify the aboveground morphological features of the target weed through the visual sensor and generate positioning data containing plant size and spatial distribution;

[0017] The calculation module is configured to calculate the voltage amplitude, frequency and action time length parameters of the initial high-frequency current according to a preset mapping relationship based on the plant size in the positioning data, execute the initial energy parameters and apply high-frequency current to the weed, and simultaneously collect the thermodynamic response signal of the weed tissue in real time.

[0018] The calibration module is configured to calibrate a plurality of high-heat response focal points in the energy action area according to the temperature rise rate distribution in the thermodynamic response signal, connect the high-heat response focal points to construct a dynamic action range boundary, and adjust the boundary shape according to the real-time thermal diffusion trend.

[0019] The division module is configured to divide the action range boundary into several independent energy regulation zones according to the thermodynamic gradient characteristics, generate a partition energy compensation coefficient based on the thermal damage degree and energy accumulation state of each regulation zone, and differentially adjust the high-frequency current output according to the partition energy compensation coefficient until the thermodynamic response signal meets the global inactivation condition.

[0020] The above-mentioned scheme of the present application at least has the following beneficial effects:

[0021] In terms of precise weeding efficiency, the aboveground morphological features of the target weed are identified through the visual sensor to generate precise positioning data covering plant size and spatial distribution information. Based on this, the initial high-frequency current parameters are calculated according to a preset mapping relationship to ensure that energy can be accurately applied to the weed. In the subsequent process, the high-heat response focal points are calibrated according to the temperature rise rate distribution in the thermodynamic response signal, and a dynamic action range boundary is constructed, which enables the energy action area to closely match the actual position of the weed and the thermal diffusion trend, realizes precise attack on the weed, effectively avoids unnecessary interference and damage to the surrounding crops and soil environment, and improves the precision of weeding operation. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is a flowchart of the energy output dynamic control method of the high-frequency electric shock weeding device provided by the embodiments of the present application.

[0023] Figure 2 is a schematic diagram of the energy output dynamic control system of the high-frequency electric shock weeding device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0024] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0025] As shown in Figure 1 the embodiments of the present application propose an energy output dynamic control method of a high-frequency electric shock weeding device, the method comprising the following steps:

[0026] Step 1, identifying the aboveground morphological characteristics of the target weeds through a visual sensor to generate positioning data containing plant size and spatial distribution;

[0027] Step 2, calculating the voltage amplitude, frequency and action time length parameters of the initial high-frequency current according to the preset mapping relationship based on the plant size in the positioning data;

[0028] Step 3, executing the initial energy parameters and applying high-frequency current to the weeds, while collecting the thermodynamic response signals of the weed tissues in real time;

[0029] Step 4, calibrating multiple high-heat response foci in the energy action area according to the temperature rise rate distribution in the thermodynamic response signals;

[0030] Step 5, connecting the high-heat response foci to construct the dynamic action range boundary and adjusting the boundary morphology according to the real-time heat diffusion trend;

[0031] Step 6, dividing the action range boundary into several independent energy regulation zones according to the thermodynamic gradient characteristics;

[0032] Step 7, generating a partition energy compensation coefficient based on the thermal damage degree and energy accumulation state of each regulation zone;

[0033] Step 8, differentially adjusting the high-frequency current output according to the partition energy compensation coefficient difference until the thermodynamic response signals meet the global inactivation condition.

[0034] In this embodiment of the invention, weed morphology and distribution are locked by visual recognition. Step 2 matches the initial energy parameters based on plant size to ensure that energy output is focused on weeds from the source, avoiding ineffective effects on surrounding crops or soil and reducing environmental interference. Thermodynamic response signals are collected in real time. Steps 4-5 calibrate the high thermal response focus and dynamically adjust the boundary of the action range so that the energy action area adapts in real time with the thermal diffusion of weeds, ensuring that different parts such as the core area of ​​tall weeds and the edge of the canopy can be effectively covered, solving the problem of incomplete local inactivation under fixed parameters. Independent control zones are divided according to thermodynamic gradients, and compensation coefficients are generated based on the degree of thermal damage and energy accumulation state of each region to achieve differentiated energy output. Energy is increased in areas with insufficient thermal response, and output is reduced in areas that have reached the standard, avoiding excessive energy consumption and significantly improving the weeding effect per unit energy consumption. By repeatedly executing the "monitoring-analysis-adjustment" closed loop, the compensation coefficient is continuously updated and the output is dynamically optimized until the conditions for full-area inactivation are met, ensuring that the weeding target can still be stably achieved under complex scenarios such as differences in weed growth status and environmental interference, improving the adaptability and reliability of the equipment.

[0035] In a preferred embodiment of the present invention, step 1 includes:

[0036] Step 11: Simultaneously capture left-view and right-view images of the target area using a calibrated binocular vision sensor;

[0037] Step 12: Perform vegetation spectral response enhancement processing on the left and right view images respectively, segment the weed pixel region and generate a binary contour mask;

[0038] Step 13: Extract the feature matching point set based on the binary contour mask, and generate a depth mapping map by calculating the binocular disparity;

[0039] Step 14: Convert the depth mapping map into a three-dimensional spatial point cloud model and fit it to the ground reference plane;

[0040] Step 15: Calculate the vertical distance from the highest point of each connected domain in the point cloud model to the ground reference plane as the plant height, and the minimum circumcircle diameter of the point cloud projected onto the horizontal plane as the canopy diameter.

[0041] Step 16: Using the coordinates of the centroid at the bottom of the point cloud connected domain as the plant position reference, output structured positioning data including three-dimensional coordinates, plant height, and canopy diameter.

[0042] In this embodiment of the invention, the specific implementation process of step 11 is as follows:

[0043] First, the binocular vision sensor is calibrated: a standard checkerboard calibration board is placed within the target shooting area, ensuring that the calibration board completely covers the sensor's field of view; the left and right cameras are controlled to capture at least 10 sets of images from different angles and distances onto the calibration board. The pixel coordinates of the checkerboard corner points in the calibration analysis images are compared with the actual physical coordinates of the calibration board. Based on these coordinate differences, the optical distortion parameters (such as radial distortion and tangential distortion) of the two cameras are calculated, and a correction matrix is ​​generated to eliminate distortion in subsequent images. Simultaneously, by analyzing the positional relationship of corresponding corner points in the left and right images, the relative positional parameters of the two cameras (such as baseline length, horizontal skew angle, and vertical skew angle) are determined to ensure that the imaging planes of the left and right cameras are precisely aligned in three-dimensional space. After calibration, the sensor's hardware synchronization trigger function is enabled, and the shutters of the left and right cameras are opened simultaneously using the same pulse signal. The target farmland area is captured within the same exposure time (e.g., 1 / 1000 second), obtaining left-view and right-view images that are completely synchronized in time and scene.

[0044] The specific implementation process of step 12 above is as follows:

[0045] Vegetation spectral response enhancement processing was performed on the left and right view images respectively: First, the color image was converted into multispectral components (such as RGB three channels and near-infrared channels). By calculating the ratio of the near-infrared channel to the red channel (i.e., the principle of NDVI vegetation index), the signal of the weed area was enhanced. Weeds contain chlorophyll, have high near-infrared reflectivity and low red reflectivity, and this ratio is significantly higher than that of soil (low near-infrared reflectivity) and crop residues (no chlorophyll, ratio close to that of soil). Histogram equalization was performed on the calculated ratio image to stretch the gray value range, making the gray value difference between weeds and the background more obvious. Then, weed pixel segmentation was performed: pixel samples of typical weed areas and non-weed areas were randomly selected from the enhanced image, and the gray value distribution of the two types of samples was statistically analyzed to determine the gray value threshold for distinguishing weeds from the background (e.g., the gray value of the weed area is higher than the threshold, and the gray value of the non-weed area is lower than the threshold). The image is judged pixel by pixel based on the threshold. Pixels with gray values ​​higher than the threshold are marked as "weed pixels" (assigned a value of 255, white), and pixels with gray values ​​lower than the threshold are marked as "non-weed pixels" (assigned a value of 0, black). Finally, a binary contour mask that can completely outline the contour of weeds is generated.

[0046] The specific implementation process of step 13 above is as follows:

[0047] Extracting Feature Matching Points from a Binary Contour Mask: Within the weed pixel region of the left-view image, edge detection algorithms (such as the Canny operator) are used to identify the edge lines of the weed contour. Points with abrupt curvature changes (such as corners) and abrupt grayscale value changes (such as texture boundaries) are then selected from these edge lines. These points are used as feature points in the left image. The same method is used to extract feature points within the weed pixel region of the right-view image. Local feature descriptors (such as feature vectors based on the grayscale distribution of surrounding pixels) are calculated for the feature points in both the left and right images. By calculating the similarity between the descriptors of the feature points in the left and right images (e.g., the smaller the vector distance, the higher the similarity), one-to-one matching feature point pairs are obtained (e.g., feature point A in the left image corresponds to feature point A' in the right image). Based on the principle of binocular vision, given the baseline lengths (physical distance obtained through calibration, e.g., 10 cm) and focal lengths (optical parameters obtained through calibration, e.g., 5 mm) of the left and right cameras, the horizontal pixel coordinate difference (i.e., parallax, e.g., the left image has an x-coordinate of 100, the right image has an x-coordinate of 80, and the parallax is 20 pixels) of each pair of feature points in the left and right images is calculated. Then, combined with the pixel size (e.g., 1 pixel corresponds to 0.01 mm), the parallax is converted into actual depth: distance from feature point to camera = baseline length × focal length ÷ (parallax × pixel size), ultimately generating a depth map corresponding to the actual depth of each pixel position.

[0048] The specific implementation process of step 14 above is as follows:

[0049] The depth map is converted into a 3D point cloud model: For each pixel in the depth map, its 2D coordinates (u, v, i.e., horizontal and vertical pixel indices) and depth value d (distance from the feature point to the camera) in the image are known. Combining the principal point coordinates (image center pixel coordinates, such as u0 = 320, v0 = 240) and focal length f (e.g., 5 mm) from the camera's intrinsic parameters, the corresponding 3D spatial coordinates of the pixel are calculated through coordinate transformation: X = (u - u0) × d ÷ f, Y = (v - v0) × d ÷ f, Z = d (Z-axis is the direction of the camera's optical axis). Traversing all pixels in the image, the calculated 3D coordinates (X, Y, Z) are summarized to form a 3D point cloud model containing all objects (weeds, soil, crops) in the target area.

[0050] The specific process of fitting the ground reference plane is as follows:

[0051] First, all points in the 3D point cloud model are traversed, and the Z-coordinate (height value along the camera's optical axis) of each point is extracted. All Z-coordinates are then sorted in ascending order. Based on the sorting, points with Z-values ​​in the lowest 10% range are selected as initial candidate points (these points, due to their low height, are likely to correspond to farmland soil surfaces). Next, the spatial continuity of these candidate points is checked: the 3D distance between any two candidate points is calculated. If the distance is less than a preset threshold (e.g., 10 cm), they are considered continuously distributed. Isolated points whose distances to other candidate points are greater than the threshold are removed (these points may be small stones, soil protrusions, or other non-ground features), ultimately yielding a preliminary set of ground candidate points.

[0052] Remove outliers using the RANSAC algorithm:

[0053] Three non-collinear points are randomly selected from the candidate ground point set (these three points define a plane). These three points are used as initial samples to fit a temporary plane. The perpendicular distance from all other points in the candidate ground point set to this temporary plane is calculated, and a distance threshold (e.g., 2 cm) is set. If the distance from a point to the plane is less than the threshold, it is determined as an "interior point" (meeting the characteristics of a ground plane); if the distance is greater than the threshold, it is determined as an "outer point" (possibly interference points such as stones in the soil or weed roots). The number of interior points corresponding to the current temporary plane is recorded. The process of random sampling, plane fitting, and interior point counting is repeated (the number of iterations is usually set to 50-100 times to ensure sufficient coverage of sample combinations). Among all iteration results, the temporary plane with the most interior points is selected as the optimal initial plane. All interior points corresponding to this plane are retained, and all outer points are removed, resulting in the purified ground point set.

[0054] Based on the RANSAC-filtered set of interior points (i.e., the cleaned ground points), the final ground reference plane is fitted by minimizing the sum of squared distances from all interior points to the plane.

[0055] Traverse each point in the inlier set, calculate the perpendicular distance from that point to the plane to be fitted, and sum the squares of all distances to obtain the total error. Adjust the parameters of the plane (a, b, c, d) to gradually reduce the total error until the total error reaches its minimum value (i.e., the overall fit between the plane and all inlier points is the highest). The equation formed by the parameter combination (a, b, c, d) at this point is the equation of the ground reference plane.

[0056] The equation of the ground reference plane is ax + by + cz + d = 0, where the three parameters a, b, and c together constitute the plane's normal vector (a vector perpendicular to the plane), i.e., the normal vector is (a, b, c). The direction of the normal vector determines the plane's tilt angle in three-dimensional space. For example, if a = 0, b = 0, and c = 1, the normal vector is perpendicular to the horizontal plane, and the plane is horizontal (corresponding to flat farmland). If a or b is not 0, the plane is tilted (corresponding to sloping farmland). d represents the vertical distance of the plane from the origin. In short, a, b, and c determine the plane's "orientation" (whether it is tilted and the tilt angle), and d determines the plane's "position" in three-dimensional space (distance from the origin and orientation). Together, they define the specific shape of the ground in three-dimensional space.

[0057] The specific implementation process of step 15 above is as follows:

[0058] Set a distance threshold between points (e.g., 5 mm), and identify connected components using the region growing method. Starting from any unmarked point, group all adjacent points that are less than the threshold distance into the same region. Repeat this process until all points are marked. Each region is a connected component (representing a single weed or a clump of dense weeds). For each connected component, iterate through the Z-coordinates of all points within it, and record the point corresponding to the maximum Z-value as the highest point of the weed. Calculate the difference between the Z-value of the highest point and the Z-value of the ground reference plane (by substituting the X and Y coordinates of the highest point into the equation of the ground plane to obtain the corresponding Z-value). This difference is the plant height (e.g., if the highest point Z = 20 cm, the corresponding ground Z = 2 cm, and the plant height is 18 cm). When calculating the canopy diameter, ignore the Z-coordinates of all points in the connected component, and only retain the X and Y coordinates to form a set of projection points on a two-dimensional plane. Use the minimum circumcircle algorithm: first find the two points farthest apart in the projection point set as the initial circle diameter, then check if all points are inside the circle. If there are external points, redraw the circle with the point and the farthest point on the circle as the new diameter. Repeat the iteration until all points are inside the circle. At this point, the diameter of the circle is the canopy diameter.

[0059] The specific implementation process of step 16 above is as follows:

[0060] For each connected component of the point cloud, points whose Z-values ​​are close to the ground reference plane are selected as the bottom point set (e.g., points whose Z-values ​​differ from the ground reference Z-value by less than 2 cm). The centroid coordinates of the bottom point set are calculated: the X-coordinates of all bottom points are added together and divided by the number of points to obtain the centroid X-coordinate; similarly, the centroid Y-coordinate and Z-coordinate (close to the ground reference Z-value) are calculated. These centroid coordinates are the position reference of the weed in three-dimensional space (e.g., X = 1.2 m, Y = 0.8 m, Z = 0.02 m). Finally, the location data is integrated: a data entry is created for each connected component (weed), containing the three-dimensional centroid coordinates (X, Y, Z), the plant height calculated in step 15 (e.g., 18 cm), and the canopy diameter (e.g., 5 cm). This entry is output in a structured table format, with table fields including "Weed ID," "X-coordinate," "Y-coordinate," "Z-coordinate," "Plant Height," and "Canopy Diameter," ensuring that subsequent steps can directly read this data for energy parameter calculation.

[0061] This invention employs a calibrated binocular vision sensor to capture stereo images, providing a reliable raw data foundation for subsequent 3D reconstruction and ensuring the accuracy of spatial positioning. Through vegetation spectral enhancement and binarization, it accurately segments weed pixel regions, effectively distinguishing weeds from backgrounds such as soil and crops, reducing interference from non-target areas, and laying a clean image foundation for feature extraction. Based on feature matching and disparity calculation, it generates a depth map, then constructs a 3D point cloud model and fits it to a ground benchmark, transforming 2D image information into 3D spatial data, achieving a stereoscopic reconstruction of weed morphology and overcoming the spatial limitations of planar images. Point cloud analysis accurately extracts plant height (vertical distance from the highest point) and canopy diameter (diameter of the projected circumcircle), quantifying key weed size characteristics and providing a direct basis for calculating initial energy parameters in step 2 (such as height-related voltage and diameter-related duration). Using the centroid of the connected domain as a position benchmark, it outputs structured positioning data, integrating core information such as 3D coordinates, height, and diameter, providing clear spatial coordinate guidance for defining the energy application range (such as boundary adjustment in step 5) and precise energy application, ensuring a strict match between energy output and the actual position and morphology of the weeds.

[0062] In a preferred embodiment of the present invention, step 2 includes:

[0063] Step 21: Based on the plant height value in the positioning data, refer to the preset mapping table between plant height range and voltage amplitude reference level to determine the corresponding voltage amplitude reference level; wherein, the mapping table is set so that different voltage reference levels correspond to different threshold ranges when the plant height value falls into different threshold ranges, and the higher the plant height value, the higher the voltage reference level.

[0064] Step 22: Based on the canopy diameter value in the positioning data, calculate the initial action duration according to a preset positive correlation mapping relationship between the canopy diameter value and the initial action duration; wherein, the positive correlation mapping relationship is set such that the larger the canopy diameter value, the longer the calculated initial action duration;

[0065] Step 23: Set a fixed high-frequency parameter value. This high-frequency parameter value is configured to be higher than the preset human safety frequency threshold and lower than the preset electromagnetic interference critical frequency threshold.

[0066] Step 24: Take the voltage amplitude reference level determined in step 21 as the initial voltage amplitude, take the initial action duration calculated in step 22 as the initial action duration, and combine it with the fixed high-frequency parameter value set in step 23 to generate the voltage amplitude, frequency and action duration parameter combination of the initial high-frequency current.

[0067] In this embodiment of the invention, the specific implementation process of step 21 is as follows:

[0068] Multiple weed electroshock experiments were conducted in laboratory and field environments. Common weed varieties (such as crabgrass, goosegrass, and lambsquarters) were selected and planted at varying heights to ensure coverage at typical growth stages: 0-5 cm, 5-10 cm, 10-15 cm, and over 15 cm. For each weed sample at each height, the voltage amplitude of the high-frequency current was gradually adjusted (increasing from low to high), and the lowest voltage value required to completely inactivate the weed roots was recorded. For example, tests showed that weeds 0-5 cm tall could be inactivated at 500 volts, with a maximum tolerance of 800 volts (exceeding this level would result in wasted energy); weeds 5-10 cm tall required voltages above 800 volts, but could be completely inactivated up to 1200 volts, and so on.

[0069] Based on experimental data, voltage amplitude benchmark levels were defined: the voltage range that "achieves inactivation without excessive energy consumption" was used as the corresponding level for each height interval. Level 1 corresponds to 0-5 cm weeds, with a voltage range of 500-800 volts; Level 2 corresponds to 5-10 cm weeds, with a voltage range of 800-1200 volts; Level 3 corresponds to 10-15 cm weeds, with a voltage range of 1200-1600 volts; and Level 4 corresponds to weeds over 15 cm, with a voltage range of 1600-2000 volts. The mapping table also indicated the rules for assigning interval boundaries, such as classifying 5 cm tall weeds as Level 2 (following the principle of "higher voltage, lower voltage" to ensure that taller weeds receive sufficient initial voltage).

[0070] From the structured location data output in step 16, the system reads the "plant height" field value of the weed to be processed through the data interface. For example, if the location data of a certain weed records its height as 12 cm, the system automatically stores this value temporarily in a temporary variable; it calls the mapping table lookup function to compare 12 cm with the threshold values ​​of each interval in the table one by one: first, it compares whether it falls into the 0-5 cm interval (12>5, no match); then it compares the 5-10 cm interval (12>10, no match); then it compares the 10-15 cm interval (10≤12≤15, successful match). At this time, the system locks the voltage amplitude reference level corresponding to this interval as 3, and extracts the voltage range of 1200-1600 volts corresponding to level 3 from the mapping table.

[0071] A second verification of the matching results is performed: check whether the current weed height value is abnormal (such as whether it exceeds the maximum range of the mapping table, or whether it is negative). If 12 cm is within a reasonable range (0-∞), then level 3 and the corresponding voltage range are confirmed to be valid. 1200-1600 volts is used as the reference range for the initial voltage amplitude, and is synchronously recorded in the energy parameter database and marked as "initial value to be adjusted". This provides an initial voltage reference for dynamic optimization in subsequent steps by combining thermodynamic response signals. At the same time, a log record is generated, indicating the current weed height, the matching range, the reference level and the voltage range, which is convenient for subsequent traceability and parameter optimization.

[0072] The specific implementation process of step 22 above is as follows:

[0073] Weed samples with different canopy diameters (e.g., 1 cm, 3 cm, 5 cm, 8 cm, 12 cm, etc.) were selected. Under fixed voltage amplitude and frequency conditions, the effect of different action durations on weed inactivation was tested. For example, for weeds with a canopy diameter of 1 cm, the test showed that 0.7 seconds of action was sufficient for complete inactivation; 1.1 seconds was required for weeds with a canopy diameter of 3 cm; 1.5 seconds was required for weeds with a canopy diameter of 5 cm, and so on. By analyzing the minimum inactivation time corresponding to different diameters, three core parameters were determined: "basic action duration," "time per unit diameter increase," and "maximum action duration." For example, when the canopy diameter is 0 cm (the theoretical minimum), the basic action time is set to 0.5 seconds (to ensure that even very small weeds can obtain basic energy); based on the experimental pattern that "for every 1 cm increase in diameter, the inactivation time increases by an average of 0.2 seconds", the time for increasing the unit diameter is determined to be 0.2 seconds / cm; at the same time, combined with energy consumption testing, the maximum action time is set to not exceed 5 seconds (to avoid excessive energy application to oversized canopy weeds), and these parameters are entered into the system to form a database of callable positive correlation mapping relationships.

[0074] From the structured location data output in step 16, the "canopy diameter" value of the current weed is extracted by matching data fields. For example, if the location data of a certain weed records its canopy diameter as 4 cm, the system stores this value in a dedicated variable. The extracted diameter value is then validated: it is checked whether the value is non-negative (to exclude abnormal data) and whether it is within a reasonable range that the equipment can process (e.g., 0-30 cm; exceeding this range may indicate a data error). If 4 cm meets the requirements, the system proceeds to the next calculation step; if the value is abnormal (e.g., a negative number or 50 cm), the system automatically triggers an error reporting mechanism, prompting the system to re-collect the data.

[0075] Calculate step by step according to the preset positive correlation mapping relationship:

[0076] For calculating the base duration, the preset base duration of 0.5 seconds (corresponding to a baseline value of 0 cm canopy diameter) is directly used. The current canopy diameter of 4 cm is multiplied by the duration increase per unit diameter of 0.2 seconds / cm, resulting in an incremental duration of 4 × 0.2 = 0.8 seconds (i.e., the additional duration added due to the canopy diameter exceeding 0 cm). The base duration and the incremental duration are added together, i.e., 0.5 seconds + 0.8 seconds = 1.3 seconds. 1.3 seconds is compared with the preset maximum duration of 5 seconds. Since 1.3 seconds < 5 seconds, it does not exceed the threshold, so the result is determined to be the effective initial duration of action. In the extreme case of a canopy diameter of 25 cm, the total duration is calculated to be 0.5 + 25 × 0.2 = 5.5 seconds. Since this exceeds the maximum threshold of 5 seconds, the system automatically corrects the initial duration to 5 seconds. The final determined initial duration (e.g., 1.3 seconds) is bound to the current weed ID and stored in the energy parameter table. At the same time, the calculation basis of this parameter is marked (including canopy diameter of 4 cm, base duration of 0.5 seconds, and incremental duration of 0.8 seconds). Then, this parameter is passed to step 24 and combined with the voltage amplitude and frequency parameters to form a complete set of initial energy output parameters, providing a precise time dimension control basis for the subsequent application of high-frequency current.

[0077] The specific implementation process of step 23 above is as follows:

[0078] First, two key frequency thresholds must be clearly defined: a preset human safety frequency threshold (e.g., 30kHz, below which current may pose a risk of electric shock) and a preset electromagnetic interference critical frequency threshold (e.g., 100kHz, above which may generate strong electromagnetic interference to surrounding electronic devices). When setting high-frequency parameters, it is necessary to ensure that they are within the safe range between these two thresholds. For example, through preliminary equipment testing and electromagnetic compatibility experiments, 80kHz can be determined as a stable output frequency. This frequency is higher than the human safety frequency threshold of 30kHz (to avoid safety hazards to operators) and lower than the electromagnetic interference critical frequency threshold of 100kHz (to reduce interference to other electronic devices in the farmland, such as sensors and controllers). 80kHz is set as a fixed high-frequency parameter value, serving as the frequency reference for the initial energy output of all weeds.

[0079] The specific implementation process of step 24 above is as follows:

[0080] Integrate the results of steps 21, 22, and 23 to generate an initial parameter combination: Obtain the specific voltage value corresponding to the determined voltage amplitude reference level from step 21 (e.g., 1400 volts for level 3, which can be an intermediate value within the voltage range of this level or the optimal value selected according to the equipment characteristics), as the initial voltage amplitude; directly use the initial action duration calculated in step 22 (e.g., 1.3 seconds) as the initial action duration parameter; use the fixed high-frequency parameter value set in step 23 (e.g., 80kHz) as the frequency parameter; finally, combine these three parameters into an initial high-frequency current parameter of "voltage amplitude 1400 volts, frequency 80kHz, action duration 1.3 seconds", ensuring that this parameter combination can provide basic energy output according to the plant height and canopy diameter of the weeds, while meeting safety and anti-interference requirements, and providing an initial reference standard for subsequent dynamic energy adjustment.

[0081] In this embodiment of the invention, step 21 matches the voltage amplitude benchmark level based on plant height, ensuring that the voltage output is adapted to the longitudinal growth scale of the weeds. This avoids insufficient or wasted energy due to height differences, ensuring that tall weeds receive sufficient initial voltage to reach deep tissues, while short weeds are not subjected to excessive voltage. Step 22 calculates the initial action time based on the canopy diameter, making the action time positively correlated with the lateral coverage of the weeds. Weeds with larger canopies receive a longer action time, ensuring that their wide areas can fully receive energy, thus solving the problem of incomplete inactivation in some areas under a uniform action time. Step 23 sets a fixed high-frequency parameter that is both higher than the human safety threshold to ensure operational safety and lower than the electromagnetic interference threshold to reduce the impact on surrounding equipment, achieving a balance between safety and anti-interference. Step 24 integrates the parameters to form a complete combination, ensuring that the initial energy output matches the height and canopy characteristics of the weeds while taking into account safety and anti-interference requirements, thereby improving energy utilization efficiency and targeted weed control from the source.

[0082] In a preferred embodiment of the present invention, step 3 includes:

[0083] Step 31: Based on the combination of voltage amplitude, frequency and duration parameters of the initial high-frequency current generated in step 24, control the high-frequency current generator to output a high-frequency current pulse sequence with corresponding voltage amplitude, frequency and planned duration.

[0084] Step 32: Apply the high-frequency current pulse sequence to the target weed plants through a contact electrode array, wherein the spatial arrangement of the electrode array covers the spatial distribution range of the weeds indicated by the positioning data generated in step 1.

[0085] Step 33: Throughout the application of the high-frequency current pulse sequence, an infrared thermal imager arranged in or near the gap of the electrode array is used to continuously capture a sequence of infrared thermal radiation images of the target weed plants and the surrounding area at a sampling rate higher than the frequency of the high-frequency current pulse.

[0086] Step 34: Perform time series analysis on the infrared thermal radiation image sequence, extract the continuous temperature value of each spatial pixel during the current action, and generate spatiotemporal temperature field data reflecting the real-time temperature changes inside the weed tissue.

[0087] Step 35: Output the spatiotemporal temperature field data as a signal characterizing the thermodynamic response of weed tissue.

[0088] In this embodiment of the invention, the specific implementation process of step 31 is as follows:

[0089] First, extract the key parameters of the initial high-frequency current from the parameter combination generated in step 24: voltage amplitude (e.g., 1400 volts), frequency (e.g., 80 kHz), and planned duration (e.g., 1.3 seconds). These parameters are transmitted to the control module of the high-frequency current generator via a data interface. The control module verifies the parameters, confirming that the voltage amplitude is within the device's rated output range (e.g., 500-2000 volts), the frequency conforms to the preset safety range (e.g., 30-100 kHz), and the duration does not exceed the maximum threshold (e.g., 5 seconds). After successful verification, the control module sends a command to the waveform generation unit inside the generator, setting the peak voltage of the pulse sequence to 1400 volts, the pulse repetition frequency to 80 kHz (i.e., 80,000 current pulses per second), and presets the total duration of the pulse sequence to 1.3 seconds via a timer. Before starting the generator, the system performs a no-load test to confirm that the voltage and frequency stability of the output pulses meet the error requirements (e.g., voltage fluctuation ≤ ±5%, frequency deviation ≤ ±1 kHz). After passing the test, the high-frequency current pulse sequence is officially output.

[0090] The specific implementation process of step 32 above is as follows:

[0091] Based on the generated positioning data, which shows the three-dimensional coordinates and spatial distribution of the weeds (e.g., a weed located at X = 1.2 meters, Y = 0.8 meters, with a canopy diameter of 4 centimeters), the system controls a mechanical adjustment mechanism to calibrate the spatial arrangement of the contact electrode array. The electrode array consists of multiple independently adjustable electrodes, with the spacing between each electrode set to be less than half the diameter of the weed canopy (e.g., 2 centimeters) to ensure uniform coverage of the weed canopy area. The X and Y axis positions of the electrodes are adjusted by a drive motor to align the center of the electrode array with the centroid coordinates of the weed's base, ensuring the electrodes completely cover the spatial boundary of the weeds indicated by the positioning data (e.g., from X = 1.18 meters to 1.22 meters, Y = 0.78 meters to 0.82 meters). Subsequently, the electrodes are controlled to descend vertically until they contact the top of the weed canopy or the surface of the stem. The contact pressure is adjusted via a pressure sensor (e.g., maintaining a pressure of 5-10 N to avoid damaging the plant or poor contact) to ensure effective current conduction to the weed tissue.

[0092] The specific implementation process of step 33 above is as follows:

[0093] An infrared thermal imager is fixed at a gap in the electrode array (e.g., at the center between adjacent electrodes) or approximately 5 cm from the electrode edge. The imager's lens field of view is adjusted to completely cover the weeds and the area within a 10 cm radius around them, ensuring the full energy impact range of the current is captured. Based on the frequency of the high-frequency current pulse (80 kHz) in step 31, the sampling rate of the thermal imager is set to 100 kHz (i.e., 100,000 infrared images per second), higher than the pulse frequency, ensuring at least one frame records the temperature change within each current pulse cycle. The thermal imager's temperature measurement range is set to ambient temperature to 100°C (covering the typical temperature range for weed tissue inactivation), with a temperature measurement accuracy calibrated to ±0.5°C. Simultaneously with the start of the high-frequency current pulse sequence output, the thermal imager synchronously starts continuous capture mode, storing each frame of image in the buffer with a timestamp (accurate to the microsecond level), ensuring complete temporal synchronization between the image sequence and the current action process.

[0094] The specific implementation process of step 34 above is as follows:

[0095] The infrared thermal radiation image sequence captured by the infrared thermal imager is imported into the image processing module in chronological order. First, spatial calibration is performed on each frame. A pre-set calibration plate is used to determine the correspondence between image pixels and actual spatial locations (e.g., 1 pixel corresponds to 0.1 mm of actual distance), ensuring that each pixel in the image can be mapped to the specific location of the weed tissue. Time-series analysis is then performed on the image sequence: the grayscale value of each pixel in each frame is extracted and converted to the corresponding temperature value using the thermal imager's grayscale-temperature conversion curve (obtained beforehand through blackbody calibration) (e.g., grayscale value 255 corresponds to 100℃). The temperature change of each spatial pixel is tracked throughout the entire current application period (1.3 seconds), and its temperature value is recorded chronologically (e.g., a pixel's temperature is 25℃ at 0.1 seconds and 40℃ at 0.5 seconds). Finally, spatiotemporal temperature field data containing three-dimensional information of "spatial coordinates (X, Y) - time - temperature" is generated. The data is stored in matrix form, with each row representing a pixel and each column representing the temperature value at a given time point.

[0096] The specific implementation process of step 35 above is as follows:

[0097] The spatiotemporal temperature field data generated in step 34 is standardized: sorted by time axis from the start to the end of the current application (0 to 1.3 seconds), and the spatial coordinates are remapped according to the actual location of the weed tissue (e.g., establishing a local coordinate system with the centroid of the weed as the origin) to ensure clear spatial correlation of the data. Key features in the data, such as the highest temperature, average temperature, and rate of temperature change at each time point, are extracted as core indicators of the thermodynamic response. The standardized spatiotemporal temperature field data is encapsulated into a structured signal, which includes a data header (recording sampling rate, time length, and spatial range) and a data body (temperature matrix). This signal is transmitted to subsequent steps (such as step 4) via an internal data bus as the raw input signal for analyzing the thermodynamic response of the weed tissue. At the same time, a data backup is stored for subsequent process traceability and parameter optimization.

[0098] In this embodiment of the invention, step 31 controls the high-frequency current generator to output a corresponding pulse sequence based on initial parameters, ensuring that energy is applied precisely according to preset specifications, providing a stable initial energy input for weed inactivation. Step 32 covers the spatial distribution range of weeds with an electrode array, ensuring precise matching between the current application area and the weed location, reducing energy diffusion to non-target areas. Step 33 continuously captures thermal radiation images with a high sampling rate infrared thermal imager, ensuring real-time and detailed capture of temperature changes in weeds under the influence of current, avoiding omission of key thermodynamic response features. Step 34 extracts continuous temperature values ​​of pixels through time series analysis, generating spatiotemporal temperature field data, transforming the abstract thermal response into intuitive data that can be quantified and analyzed. Step 35 outputs a thermodynamic response signal, providing a real-time basis for subsequent calibration of high thermal response focal points and dynamic adjustment of energy output, realizing the starting point of a closed loop of "energy application-monitoring-feedback," ensuring that energy regulation is based on the actual response of weeds, and improving the accuracy and efficiency of weed control.

[0099] In a preferred embodiment of the present invention, step 4, calibrating multiple high thermal response focal points within the energy action region based on the temperature rise rate distribution in the thermodynamic response signal, includes:

[0100] Step 41: Based on the spatiotemporal temperature field data output in step 35, calculate the temperature change per unit time of each spatial pixel within the energy action zone within a preset time window, and generate the corresponding temperature rise rate field data.

[0101] Step 42: Perform spatial domain analysis on the temperature rise rate field data to identify all pixels whose temperature rise rate values ​​exceed a preset first rate threshold, forming a candidate high thermal response point set.

[0102] Step 43: Perform spatial clustering on the candidate high thermal response point set, and aggregate the candidate points that meet the preset spatial proximity conditions into several independent high thermal response clusters.

[0103] Step 44: Calculate the average temperature rise rate of all points in each high thermal response cluster, and select clusters whose average value exceeds the preset second rate threshold as effective high thermal response clusters.

[0104] Step 45: Determine the spatial center coordinates of each effective high thermal response cluster, and mark the center coordinates as a high thermal response focus within the energy action zone, and output the set of coordinates of all marked high thermal response focuses.

[0105] In this embodiment of the invention, the specific implementation process of step 41 is as follows:

[0106] From the spatiotemporal temperature field data output in step 35, extract the coordinates of all spatial pixels in the energy-affected area (i.e., the weeds and surrounding area covered by the electrode array) and their corresponding time-temperature sequences (temperature values ​​of each pixel at different time points); based on the periodic characteristics of the high-frequency current pulse (e.g., pulse frequency 80kHz, period approximately 0.0125 milliseconds), set the time window to 5 pulse periods (e.g., 0.0625 milliseconds) to ensure that a complete temperature change fluctuation can be captured; for each spatial pixel, extract a continuous preset time window from the time sequence (e.g., from time t1 to time t2). (t2-t1=0.0625 milliseconds), extract the initial temperature value T1 (at time t1) and the final temperature value T2 (at time t2) within the window, and calculate the temperature change ΔT=T2-T1; divide ΔT by the duration of the time window (0.0625 milliseconds) to obtain the temperature change per unit time of the pixel within the window (i.e., the temperature rise rate, in ℃ / millisecond); repeat the above calculation for all pixels in the entire energy action area, and store the temperature rise rate of each point according to its spatial coordinates to generate temperature rise rate field data covering the entire action area (each coordinate point corresponds to a temperature rise rate value).

[0107] The specific implementation process of step 42 above is as follows:

[0108] A preset first rate threshold is established: based on the critical temperature rise rate at which "irreversible thermal damage begins to occur in weed tissue" observed in previous experiments (e.g., when the temperature rise rate exceeds 0.5℃ / ms, weed cells begin to coagulate), the first rate threshold is set to 0.5℃ / ms. Each spatial pixel in the temperature rise rate field data is iterated through, and its temperature rise rate value is compared with the first rate threshold one by one. If the temperature rise rate of a pixel is >0.5℃ / ms, then that pixel is determined as a "candidate high thermal response point," and its spatial coordinates are recorded (e.g., X = 10.2cm, Y = 8.5cm); if it is ≤0.5℃ / ms, then that point is excluded. The coordinates of all qualified candidate high thermal response points are summarized to form a candidate high thermal response point set, ensuring that the point set only contains areas that respond violently to the current (rapid temperature rise).

[0109] The specific implementation process of step 43 above is as follows:

[0110] Setting spatial proximity conditions: Based on the continuity of weed tissue (such as the heat conduction distance between adjacent cells), a preset spatial distance threshold of 0.3cm is set (i.e., when the straight-line distance between two points is ≤0.3cm, they are considered spatially proximate). An unmarked candidate point is randomly selected from the point set as the initial core point. Centered on this point, all other candidate points in the point set with a distance ≤0.3cm from it are searched. These points are grouped into the same temporary cluster with the core point and marked as "processed". For points in the temporary cluster that are not core points, the above search process is repeated (centered on this point, unmarked points with a distance ≤0.3cm are included) until the cluster no longer expands. From the remaining unmarked candidate points, a new core point is selected, and the steps are repeated until all candidate points are marked and grouped into the corresponding clusters, ultimately forming several independent high heat response clusters. Each cluster represents a continuous high heat response area (such as the growth point of weeds or a dense vascular bundle area).

[0111] The specific implementation process of step 44 above is as follows:

[0112] Calculate the average temperature rise rate for each cluster: For each high thermal response cluster, extract the temperature rise rate values ​​of all candidate points, add these values ​​together and divide by the number of points in the cluster (i.e., the total number of points) to obtain the average temperature rise rate of the cluster. For example, a cluster contains 5 points with rates of 0.6, 0.7, 0.8, 0.7, and 0.6 °C / ms, respectively, and an average rate of (0.6 + 0.7 + 0.8 + 0.7 + 0.6) ÷ 5 = 0.68 °C / ms. To ensure that the cluster as a whole has a significant high thermal response (excluding sporadic clusters of high-rate points), a second rate threshold is set at 0.6 °C / ms (higher than the first threshold, reflecting the high thermal characteristics at the cluster level). The average temperature rise rate of each cluster is compared with the second threshold. If the average rate is > 0.6 °C / ms, it is determined to be an "effective high thermal response cluster"; if it is ≤ 0.6 °C / ms, it is determined to be an invalid cluster and is removed. For example, the cluster with an average rate of 0.68 °C / ms is retained, while the cluster with an average rate of 0.55 °C / ms is removed.

[0113] The specific implementation process of step 45 above is as follows:

[0114] For each effective high-thermal-response cluster, extract the spatial coordinates (X1, Y1), (X2, Y2), ..., (Xn, Yn) of all points within it. Calculate the average value of the X coordinates (X center = (X1 + X2 + ... + Xn) ÷ n) and the average value of the Y coordinates (Y center = (Y1 + Y2 + ... + Yn) ÷ n) to obtain the spatial center coordinates (X center, Y center) of the cluster. Define the spatial center coordinates of each effective cluster as a "high-thermal-response focus," representing the core location of the cluster (i.e., the area with the most concentrated thermal response in the weed tissue). Summarize the center coordinates of all effective clusters to form a set of high-thermal-response focus coordinates (e.g., [(10.3cm, 8.6cm), (12.1cm, 9.2cm), ...]), and output this set to step 5 as the core reference point for constructing the dynamic action range boundary.

[0115] Step 41 of this invention calculates the global temperature rise rate field, quantifying the thermal dynamic changes of weed tissue into spatial distribution data, providing a comprehensive basis for identifying high thermal response areas and avoiding omission of key reaction areas; Step 42 filters candidate points through a first rate threshold, accurately locking areas sensitive to current action and with rapid temperature rise, eliminating low-response background interference, and focusing on potential key inactivation areas; Step 43 spatial clustering aggregates neighboring candidate points into clusters, eliminating the influence of single-point noise, highlighting continuously distributed high thermal response areas, and better reflecting the actual thermal response characteristics of weed tissue; Step 44 filters effective clusters with a second rate threshold, ensuring that the retained clusters have a significant overall high thermal response, further improving the reliability of focus calibration and avoiding interference from atypical areas; Step 45 the high thermal response focus determined in step 45 accurately locates the core area with the most intense thermal response in weed tissue, providing a clear target for subsequent dynamic adjustment of the energy action range, making energy regulation more focused on key inactivation points, and improving weeding efficiency and accuracy.

[0116] In a preferred embodiment of the present invention, step 5 includes:

[0117] Step 51: Based on the set of all calibrated high thermal response focus coordinates output in step 45, perform a spatial topology connection operation to connect adjacent focuses that meet the preset distance threshold condition in pairs to form an initial polygonal boundary network.

[0118] Step 52: Perform boundary smoothing on the initial polygon boundary network to eliminate local sharp corners and generate a continuous closed initial scope boundary profile.

[0119] Step 53: Based on the real-time updated spatiotemporal temperature field data output in step 35, calculate the thermal diffusion rate field data at the current moment within the energy action zone; the thermal diffusion rate field data reflects the spatial gradient characteristics of heat propagation from the high-temperature zone to the low-temperature zone.

[0120] Step 54: Based on the thermal diffusion rate field data, predict the anisotropic thermal diffusion trend within a preset distance from the outer edge of the initial action range boundary contour; wherein, the region with a high thermal diffusion rate is predicted as the boundary expansion direction, and the region with a low thermal diffusion rate is predicted as the boundary maintenance or contraction direction.

[0121] Step 55: Based on the anisotropic thermal diffusion trend predicted in step 54, the initial action range boundary contour is deformed and adjusted in real time: when the predicted direction is expansion, the boundary is extended outward by a preset deformation; when the predicted direction is maintenance or contraction, the boundary is maintained or contracted inward; a dynamically updated action range boundary shape is generated, and the action range boundary shape data after dynamic adjustment at the current moment is output.

[0122] In this embodiment of the invention, the specific implementation process of step 51 is as follows:

[0123] From the set of high thermal response focus coordinates output in step 45, extract the three-dimensional spatial coordinates of all focuses (e.g., (X1, Y1, Z1), (X2, Y2, Z2)...(Xn, Yn, Zn)) and project them onto a horizontal plane (ignore the Z coordinate, retain the X and Y coordinates) for planar topology analysis. Based on the average density of the weed canopy and the focus distribution characteristics, set a distance threshold of 5 cm for adjacent focuses (i.e., when the straight-line distance between two focuses is ≤ 5 cm, they are considered "adjacent focuses"). Calculate the Euclidean distance between any two focuses in the set and compare the distance of each pair of focuses with the threshold. If the distance is ≤ 5 cm, they are considered adjacent, and the two focuses are connected pairwise by straight lines. If the distance is > 5 cm, they are not connected. After traversing all focus pairs, an initial polygonal boundary network (e.g., irregular polygonal combinations such as triangles and quadrilaterals) composed of multiple line segments is formed, ensuring that the network covers all focuses and the line segments do not intersect.

[0124] The specific implementation process of step 52 above is as follows:

[0125] For each corner (i.e., the intersection of three or more line segments) in the initial polygonal boundary network, identify local sharp corners (e.g., vertices with corner angles < 90°); for each sharp corner, take the midpoint of the two adjacent line segments constituting the corner, and calculate the midpoint coordinates (e.g., the endpoints of the line segments are A(Xa, Ya) and B(Xb, Yb), and the midpoint is M[(Xa+Xb) / 2, (Ya+Yb) / 2]); replace the original corner vertex with the midpoint of the line connecting the midpoints of the adjacent line segments, or fit the adjacent line segments with a Bézier curve to make the line segments at the corner transition smoothly (e.g., adjust the angle to 120°-150°), and repeat the above smoothing operation on the entire boundary network until all corners meet the preset smoothness requirements (e.g., corner angle ≥ 100°), and finally form a continuous and closed initial functional range boundary profile (i.e., a polygonal profile without breaks or sharp protrusions).

[0126] The specific implementation process of step 53 above is as follows:

[0127] From the real-time updated spatiotemporal temperature field data output in step 35, extract the temperature values ​​of all spatial pixels within the energy field at the current time (e.g., time t) to form the current temperature field distribution (each pixel corresponds to a temperature value T(x, y)). For any two adjacent pixels in the temperature field (e.g., pixel P(x1, y1) and pixel Q(x2, y2)), calculate the temperature difference ΔT = |T(x1, y1) - T(x2, y2)|, and measure the spatial distance d between the two points (in centimeters). The thermal diffusion rate v = ΔT / d (in °C / cm), with the rate direction pointing from the high-temperature pixel to the low-temperature pixel (i.e., the direction of heat propagation). Traverse all adjacent pixel pairs throughout the entire energy field, and store the thermal diffusion rate (including magnitude and direction) of each pixel according to its spatial coordinates to generate thermal diffusion rate field data. This data reflects the spatial gradient characteristics of heat propagation from high-heat areas (e.g., near the focus of high thermal response) to low-temperature areas (e.g., the edge of weeds or soil) in vector form.

[0128] The specific implementation process of step 54 above is as follows:

[0129] The outer edge is set to a preset distance of 2 cm, that is, a ring-shaped area extending 2 cm outward from the boundary outline, which is used as the target area for predicting the thermal diffusion trend. The thermal diffusion rate of all pixels in the area is statistically analyzed, and a rate threshold (such as 0.5℃ / cm) is set. Areas with a rate > 0.5℃ / cm are identified as "high thermal diffusion rate areas", and areas with a rate ≤ 0.5℃ / cm are identified as "low thermal diffusion rate areas". High thermal diffusion rate areas are predicted as "boundary expansion direction" because the heat spreads quickly, indicating that the weed tissue in this direction is still rapidly absorbing heat and needs to expand the energy range to cover the diffusion area. Low thermal diffusion rate areas are predicted as "boundary maintenance direction" (when the rate is close to the threshold) or "boundary contraction direction" (when the rate is much lower than the threshold).

[0130] The specific implementation process of step 55 above is as follows:

[0131] Based on the prediction results of step 54, set the boundary adjustment parameters: In the expansion direction, determine the extension amount according to the magnitude of the heat diffusion rate (e.g., for every 0.1℃ / cm increase in rate, the extension amount increases by 0.3cm, with a maximum extension amount not exceeding 1cm); in the maintenance direction, set the boundary position to remain unchanged; in the contraction direction, determine the contraction amount according to the degree of rate reduction (e.g., for every 0.1℃ / cm decrease in rate, the contraction amount increases by 0.2cm, with a maximum contraction amount not exceeding 0.5cm); traverse each vertex of the boundary contour, determine the predicted trend of the vertex's direction, if it is the expansion direction, move the vertex along the heat diffusion direction by the corresponding extension amount; if it is the maintenance direction, keep the vertex position unchanged; if it is the contraction direction, move the vertex inwards towards the contour by the corresponding contraction amount; after adjustment, reconnect all vertices to form a new closed contour, generate a dynamically updated boundary shape of the effective range, and output the boundary shape data at the current moment (including contour vertex coordinates, boundary length, etc.), ensuring that the boundary always conforms to the real-time heat diffusion state of the weed tissue.

[0132] In this embodiment of the invention, step 51 constructs an initial polygonal boundary network based on the high thermal response focus, enabling the boundary to accurately surround the core thermal response area and ensuring that the energy effect range is focused on the key inactivation zone of weeds, avoiding excessive coverage of non-target areas by the initial boundary; step 52's boundary smoothing process eliminates sharp corners, making the initial boundary more consistent with the continuous distribution characteristics of weed tissue, reducing the problem of uneven energy distribution caused by boundary irregularities; step 53 calculates the thermal diffusion rate field, capturing the spatial gradient of heat propagation in real time, providing a quantitative basis for dynamic boundary adjustment that reflects the actual trend of thermal diffusion, ensuring that the adjustment direction conforms to the real thermal behavior of weed tissue; step 54 predicts the anisotropic thermal diffusion trend, enabling targeted identification of thermal diffusion needs in each direction of the boundary, providing clear directional guidance for differentiated adjustments, and avoiding energy waste or insufficient coverage caused by indiscriminate adjustments; step 55 dynamically adjusts the boundary shape according to the predicted trend, making the effect range adapt to thermal diffusion in real time, expanding in the high thermal diffusion zone to cover the newly heated area, and contracting in the low thermal diffusion zone to reduce ineffective energy consumption, ultimately achieving dynamic optimization of the energy effect range, ensuring full-area inactivation of weeds, and significantly improving energy utilization efficiency.

[0133] In a preferred embodiment of the present invention, step 6, dividing the boundary of the effective range into several independent energy control regions according to thermodynamic gradient characteristics, includes:

[0134] Step 61: Based on the dynamically adjusted boundary shape data of the scope of action output in step 55, extract the closed area enclosed by the boundary as the basic control range.

[0135] Step 62: Based on the real-time updated spatiotemporal temperature field data output in step 35, calculate the temperature change rate gradient value and heat accumulation dose value at each spatial location point within the basic control range.

[0136] Step 63: Based on the temperature change rate gradient value and heat accumulation dose value calculated in step 62, generate a thermodynamic gradient distribution map covering the basic control range;

[0137] Step 64: Divide the thermodynamic gradient distribution map into equal gradient lines to form sub-regions with similar thermodynamic gradient characteristics;

[0138] Step 65: Define each continuous sub-region with similar thermodynamic gradient characteristics obtained in Step 64 as an independent energy control region, and output the spatial range identifier of all independent energy control regions.

[0139] In this embodiment of the invention, the specific implementation process of step 61 is as follows:

[0140] From the dynamically adjusted boundary shape data of the effective range output in step 55, read the structured data storing the boundary coordinates (usually a coordinate array, in the format [(X1, Y1), (X2, Y2), ..., (Xn, Yn)]). Extract each coordinate point in the array one by one, and record its specific values ​​on the X and Y axes (e.g., in centimeters, accurate to 0.01cm) to ensure that no coordinate points are missed or repeated; by calculating the direction of the line connecting adjacent coordinate points (e.g., comparing the offset direction of the later point relative to the previous point), determine whether the coordinate points are arranged consecutively in clockwise or counterclockwise order. For example, if the line connecting (X1, Y1) to (X2, Y2) and then to (X3, Y3) always deflects to the same side (e.g., all deflect to the left), it is determined to be an ordered arrangement; compare the values ​​of the first coordinate point (X1, Y1) and the last coordinate point (Xn, Yn) in the array, if the difference in X is ≤0.01cm and the difference in Y is ≤0.01cm (the set error threshold, allowing slight calculation errors), it is determined that the boundary ends coincide; if the difference exceeds the threshold, the data generation process in step 55 needs to be traced back to correct the boundary coordinates to eliminate gaps and ensure that the polygon is not broken.

[0141] The ray casting method is used as the core algorithm for spatial region clipping. The specific operation is as follows:

[0142] Based on the extreme values ​​of the boundary coordinates (e.g., minimum X value Xmin, maximum X value Xmax, minimum Y value Ymin, maximum Y value Ymax), determine the rectangular boundary of the area to be verified (X∈[Xmin-0.5cm, Xmax+0.5cm], Y∈[Ymin-0.5cm, Ymax+0.5cm]), ensuring coverage of the boundary and surrounding possible areas, avoiding omission of points near the boundary; within the above rectangular area, generate uniformly distributed verification points according to a preset grid spacing (e.g., 0.05cm×0.05cm, balancing accuracy and efficiency), with the coordinates of each point being... ,in, Starting from Xmin-0.5cm, increment by 0.05cm to Xmax+0.5cm; Similarly, for each grid verification point, a "point inside polygon" check is performed: draw a virtual ray from the point to the horizontal right (positive X-axis direction), and count the number of intersections between the ray and the polygon boundary. If the number of intersections is odd, the point is determined to be inside the polygon (marked as "within the basic control range"); if it is even or 0, it is determined to be outside (marked as "outside the range"). If the ray happens to pass through the boundary coordinate point or coincides with the boundary edge, the ray direction needs to be finely adjusted (e.g., offset by 0.001cm in the positive Y-axis direction) and recalculated to avoid misjudgment. After verifying all grid points, summarize all points marked as "within the basic control range" to form a complete basic control range spatial definition, ensuring that subsequent steps (such as temperature parameter calculation in step 62) are only performed on points within this range, completely excluding non-target areas such as soil and crops outside the boundary.

[0143] The specific implementation process of step 62 above is as follows:

[0144] Based on the spatial boundary of the basic control range (defined in step 61), the grid parameters are set as follows: the grid spacing is 0.1cm × 0.1cm, that is, a grid point is divided every 0.1cm along the X and Y axes to ensure that all grid points fall within the "basic control range". A unique coordinate identifier is assigned to each grid point through coordinate mapping (e.g., (X1, Y1), (X1, Y1 + 0.1cm)). From the real-time spatiotemporal temperature field data output in step 35, the temperature change rate of each grid point at the current moment is extracted (i.e., the "temperature change per unit time" calculated in step 41, in °C / ms). During extraction, the data correspondence must be strictly verified: by matching the grid point coordinates with the spatial coordinates of the temperature field data, the accuracy of the temperature change rate of each grid point is ensured (e.g., excluding mismatches caused by data transmission delays). If a grid point lacks corresponding temperature data (e.g., edge data is missing), the average of its three nearest valid grid points is used to fill the gap, avoiding data omissions. For each grid point, neighboring points are selected according to the "8-neighborhood rule": namely, adjacent grid points in eight directions: top (X, Y+0.1cm), bottom (X, Y-0.1cm), left (X-0.1cm, Y), right (X+0.1cm, Y), top left (X-0.1cm, Y+0.1cm), top right (X+0.1cm, Y+0.1cm), bottom left (X-0.1cm, Y-0.1cm), and bottom right (X+0.1cm, Y-0.1cm). For boundary grid points (e.g., points at the edge of the basic control range), if there are no neighboring grid points in a certain direction (e.g., the leftmost point has no left neighbor), that direction is discarded, and only the existing neighboring points are retained (e.g., the leftmost point retains 7 or fewer neighboring points), and the missing direction is recorded to avoid invalid calculations.

[0145] Calculate the temperature change rate difference between the current point and each neighboring point: Subtract the temperature change rate of the neighboring points from the current point's temperature change rate and take the absolute value (e.g., if the current point's rate is 0.8℃ / ms and the right-hand point's is 0.6℃ / ms, the difference is |0.8-0.6|=0.2℃ / ms). Repeat this operation for all valid neighboring points to obtain a set of difference data (e.g., 8 or 7). Summarize the difference data of all neighboring points and calculate the comprehensive gradient value using the "arithmetic mean method": Add all the differences and divide by the number of valid neighboring points (e.g., divide by 8 for 8 neighboring points, divide by 7 for 7), to obtain the "temperature change rate gradient value" of that grid point. For example, if the differences between a point and its 8 neighboring points are 0.2, 0.3, 0.1, 0.2, 0.4, 0.3, 0.2, and 0.1℃ / ms, the sum is 1.8℃ / ms, and the gradient value is 1.8÷8=0.225℃ / ms. The higher the gradient value, the more significant the difference in thermal response between the point and the surrounding area (such as the difference between the core area and the edge area of ​​weeds). From the spatiotemporal temperature field data in step 35, extract the complete temperature record of the current grid point from the start of the high-frequency current application (denoted as the initial time t0) to the current time tn, forming a historical temperature sequence: T(t0), T(t1), T(t2), ..., T(tn), where each temperature value corresponds to a time point (such as t0 = 0ms, t1 = 0.01ms, t2 = 0.02ms, ..., tn = the current time).

[0146] The continuity of the time series needs to be verified during extraction: if there are missing time points (e.g., no temperature data in a 0.01ms interval), linear interpolation is used to supplement them (e.g., estimating the temperature of tk+1 based on the temperature values ​​of the previous time tk and the next time tk+2) to ensure that the series is not broken; the time interval Δt = 0.01ms is set (to match the temperature sampling frequency and ensure coverage of every temperature change detail). The total number of intervals n in the historical temperature series is counted: that is, there are n intervals from t0 to tn (n = (tn-t0) / 0.01ms).

[0147] Add the temperature values ​​T(ti) (i = 0 to n-1) of each time interval to get the total temperature sum ΣT = T(t0) + T(t1) + ... + T(tn-1); then multiply the total temperature sum by the time interval Δt, that is, the cumulative heat dose = ΣT × 0.01ms. For example, if the sum of the temperature values ​​of 10 intervals is 50℃, then the dose = 50 × 0.01 = 0.5℃·ms. This value reflects the total heat intensity received at that point from the application of current to the current moment. If the cumulative heat dose value is negative (which is impossible) or far exceeds the average level of similar grid points (e.g., more than 3 times the average), then the temperature sequence extraction process is traced back to check for data errors (e.g., abnormal temperature values), and recalculated to ensure that the dose value accurately reflects the actual heat accumulation. All grid points within the basic control range are traversed in a "row-by-row, column-by-column" order: starting from the leftmost column with the smallest X value, each grid point is processed sequentially along the Y-axis. After completing one column, the process moves to the next column until all points are processed. For each grid point, the calculated "temperature change rate gradient value" and "cumulative heat dose value" are bound to the coordinate identifier of that point and stored as structured data (e.g., each entry contains coordinates (X, Y), gradient value, and dose value) to ensure that subsequent steps (e.g., generating a thermodynamic gradient distribution map in step 63) can directly call the corresponding parameters. After the traversal is complete, output a dataset containing all grid point parameters as input for step 63.

[0148] The specific implementation process of step 63 above is as follows:

[0149] A two-dimensional coordinate system (consistent with the spatial coordinates of the basic control range) is established, and the "temperature change rate gradient value" and "heat accumulation dose value" of each spatial point calculated in step 62 are mapped to the corresponding positions in the coordinate system. Data visualization mapping rules are adopted: for example, the temperature change rate gradient value is represented by color intensity (dark color represents high gradient, light color represents low gradient), and the heat accumulation dose value is represented by transparency (high transparency represents low dose, low transparency represents high dose). By superimposing the features of the two parameters, a thermodynamic gradient distribution map covering the entire basic control range is generated. The visual features (color + transparency) of each position in the map intuitively reflect the comprehensive thermodynamic gradient characteristics of that point (the degree of difference in thermal response and the total heat accumulation state). The distribution map is data-verified to ensure that the gradient features of each spatial point are accurately mapped without data loss or mismatch (such as whether the gradient values ​​of edge points are reasonable).

[0150] The specific implementation process of step 64 above is as follows:

[0151] Based on the thermodynamic gradient distribution map generated in step 63, extract the temperature change rate gradient values ​​and cumulative thermal dose values ​​for all spatial points, and determine the value ranges of the two parameters (e.g., gradient value 0.1-1.0℃ / ms, dose value 5-50℃·ms). Divide the data into several continuous intervals according to the distribution characteristics of the gradient and dose values ​​(e.g., intervals of 0.2℃ / ms for gradient values ​​and intervals of 10℃·ms for dose values), ensuring that the parameter values ​​within each interval are relatively similar (i.e., "similar thermodynamic gradient characteristics"). Starting from any unmarked point, check whether the gradient and dose values ​​of its adjacent points fall into the same interval (e.g., if the gradient at the current point is 0.3-0.5℃ / ms and the dose is 10-20℃·ms, adjacent points must satisfy the same interval). If they satisfy this condition, the adjacent points are included in the same sub-region. Repeat this expansion until no adjacent points meet the conditions, forming a sub-region. Repeat this process for all unmarked points, ultimately obtaining multiple independent continuous sub-regions with similar internal parameter characteristics.

[0152] The specific implementation process of step 65 above is as follows:

[0153] For each sub-region divided in step 64, its spatial boundary (such as the set of coordinate points outside the sub-region) is determined by a boundary extraction algorithm, and the continuity of the region is verified (i.e., there are no gaps or breaks inside, and all points are connected). A unique identifier is assigned to each continuous sub-region (such as "Control Zone 1" and "Control Zone 2"), and its spatial range characteristics are recorded: for example, the minimum / maximum X and Y coordinates of all points in the sub-region are used to represent the rectangular boundary range, or the boundary coordinate sequence of the sub-region is directly recorded. The identifier is bound to the spatial range characteristics to form an "Independent Energy Control Zone List", in which each entry contains the control zone ID, boundary coordinate range, and core thermodynamic gradient characteristics (such as average gradient value and average dose value), ensuring that subsequent steps can accurately locate the spatial position and characteristics of each control zone through this list. Finally, this list is output as the result of step 6.

[0154] Step 61 of this invention extracts the closed basic control range, ensuring that energy control focuses only on the dynamically adjusted effective area, avoiding ineffective coverage of non-target areas outside the boundary, and spatially locking the control target. Step 62 calculates the temperature change rate gradient value and heat accumulation dose value at each point, quantifying the thermal dynamic differences of weed tissues (such as the speed of thermal response and the amount of total heat accumulation) into analyzable parameters, providing detailed thermodynamic characteristic data support for subsequent zoning. Step 63 generates a thermodynamic gradient distribution map, intuitively presenting the spatial distribution of thermal characteristics within the basic control range, making the thermodynamic differences between different regions clear at a glance, and providing a visual reference for zoning. Step 64 divides similar characteristic sub-regions by isogradation lines, aggregating regions with similar thermodynamic characteristics, ensuring that the thermal response pattern of weeds is consistent within each sub-region; Step 65 defines independent energy control zones and outputs identifiers, enabling subsequent energy adjustments to be performed according to regional differences, increasing energy in areas with insufficient thermal response, and reducing output in areas with sufficient heat accumulation, achieving "on-demand energy application," significantly improving the accuracy and efficiency of energy control, and avoiding energy waste or incomplete local inactivation.

[0155] In a preferred embodiment of the present invention, step 7 includes:

[0156] Step 71: Obtain the cumulative thermal dose value of all points in each independent energy control zone at the current time; calculate the average cumulative thermal dose of all points in the control zone as the current energy accumulation state characterization value of the control zone; obtain the temperature value of all points in the control zone at the current time; and count the proportion of pixels in the control zone whose temperature value reaches the critical temperature threshold for tissue inactivation as the current thermal damage characterization value of the control zone.

[0157] Step 72: Compare the current energy accumulation state characterization value with the preset target accumulated energy value of the control area, and calculate the energy accumulation difference;

[0158] Step 73: Compare the current thermal damage degree characterization value with the preset target thermal damage degree threshold for the control area, and calculate the thermal damage degree difference;

[0159] Step 74: Calculate the energy compensation demand intensity of the control zone based on the energy accumulation difference and the thermal damage degree difference;

[0160] Step 75: Normalize the energy compensation demand intensity to generate the partition energy compensation coefficient for the control zone; generate the partition energy compensation coefficient for each independent energy control zone.

[0161] In this embodiment of the invention, the specific implementation process of step 71 is as follows:

[0162] Based on the independent energy control zone spatial range identifiers output in step 65 (such as the boundary coordinates of each control zone), accurately locate the spatial range of each control zone. For each control zone, extract the "cumulative heat dose value" and "current temperature value" of all spatial points (by grid points) within that range from the structured data generated in step 62, excluding points outside the control zone and invalid data (such as points with abnormal temperature values, which are filtered by a preset threshold, such as temperatures <0℃ or >100℃ being determined as invalid and replaced with the average of adjacent valid points); summarize the extracted cumulative heat dose values: count the total number of valid points within the control zone (denoted as N), and sum the cumulative heat dose values ​​of all valid points to obtain the total (denoted as S). Calculate the average value: Current energy accumulation state characterization value = S ÷ N. For example, if a control zone has 100 effective points with a total dose of 500℃·ms, then the average value = 500 ÷ 100 = 5℃·ms, which reflects the overall heat accumulation level of the area. A preset critical temperature threshold for tissue inactivation is used (based on experimental data, such as a critical temperature of 60℃ for complete inactivation of weed cells). The temperature values ​​within the control zone are statistically analyzed: the number of effective points with a temperature ≥ 60℃ is counted (denoted as M), and the proportion is calculated: current thermal damage level characterization value = M ÷ N × 100%. For example, if 65 out of 100 effective points reach 60℃, then the proportion = 65 ÷ 100 × 100% = 65%, which reflects the percentage of weed tissue in the area that has been thermally damaged.

[0163] The specific implementation process of step 72 above is as follows:

[0164] Preset a target cumulative energy value for each control zone: Based on the thermodynamic gradient characteristics of the control zone (such as the gradient distribution map in step 63) and the weed type (the weed species associated with the previous location data), set differentiated target values. For example, the target value for control zones with slower thermal response (low gradient value) is set to 8℃·ms, and for areas with faster thermal response, it is set to 6℃·ms, ensuring that the target matches the regional characteristics; compare the "current energy accumulation state characterization value" obtained in step 71 with the preset target cumulative energy value and calculate the difference: Energy accumulation difference = target cumulative energy value - current average energy accumulation. For example, if the target value is 8℃·ms and the current average value is 5℃·ms, then the difference = 8 - 5 = 3℃·ms (a positive value indicates insufficient energy accumulation, which needs to be supplemented; a negative value indicates excessive accumulation).

[0165] The specific implementation process of step 73 above is as follows:

[0166] A target thermal damage threshold is preset for each control zone (based on weed control requirements, such as needing 90% of weed tissue to reach the inactivation temperature). The threshold can be adjusted according to the importance of the control zone, such as setting the threshold to 95% for the core high-heat response zone and 85% for the edge zone. The "current thermal damage degree characterization value" (proportion) obtained in step 71 is compared with the target threshold, and the difference is calculated: thermal damage degree difference = target threshold - current thermal damage proportion. For example, if the target threshold is 90% and the current proportion is 65%, then the difference = 90% - 65% = 25% (a positive value indicates insufficient damage, which needs to be strengthened; a negative value indicates excessive damage).

[0167] The specific implementation process of step 74 above is as follows:

[0168] Based on the weighted impacts of energy accumulation and thermal damage on weed control effectiveness, pre-set weighting coefficients for both (e.g., a weight of 0.6 for the energy accumulation difference and 0.4 for the thermal damage difference, determined based on their contribution to the inactivation effect in experiments). Multiply the energy accumulation difference from step 72 and the thermal damage difference from step 73 by their respective weights, then sum them to obtain the demand intensity: Energy compensation demand intensity = (Energy accumulation difference × 0.6) + (Temperature damage difference × 0.4). For example, if the energy difference is 3℃·ms and the thermal damage difference is 25%, then the demand intensity = 3 × 0.6 + 25% × 0.4 = 1.8 + 0.1 = 1.9 (the larger the value, the stronger the compensation demand). If a difference is negative (e.g., excessive energy), that part is calculated as 0 to avoid reverse compensation.

[0169] The specific implementation process of step 75 above is as follows:

[0170] Collect the energy compensation demand intensity values ​​of all independent energy regulation zones, and find the maximum value (denoted as Max) and minimum value (denoted as Min). If all demand intensities are 0, the compensation coefficient is set to 0; otherwise, normalize the demand intensity of each regulation zone: Zone energy compensation coefficient = (current demand intensity - Min) ÷ (Max - Min). For example, if the demand intensity of a regulation zone is 1.9, and Max = 2.0 and Min = 0.5 in all regions, then the coefficient = (1.9 - 0.5) ÷ (2.0 - 0.5) = 1.4 ÷ 1.5 ≈ 0.93 (the coefficient ranges from 0 to 1, the closer to 1, the stronger the compensation demand); bind the normalized coefficient of each regulation zone to its spatial range identifier to generate a "regulation zone ID - compensation coefficient" correspondence table to ensure that subsequent energy adjustments can accurately match the compensation intensity according to the region. Output this table as the input basis for step 8. Through the above process, step 7 realizes the quantitative assessment of the energy state of each regulation zone and the accurate calculation of compensation demand.

[0171] In a preferred embodiment of the present invention, step 8 includes:

[0172] Step 81: Based on the regional energy compensation coefficients of each independent energy regulation zone generated in step 75, establish a mapping table between the spatial identifiers of the regulation zones and the compensation coefficients.

[0173] Step 82: Based on the mapping table, analyze the required high-frequency current output increment ratio for each independent energy control zone and generate the corresponding electrode control instruction set;

[0174] Step 83: Execute the electrode control instruction set to drive the high-frequency current generator to implement differentiated current output for each independent energy regulation zone divided in step 6.

[0175] Step 84: During the differentiated current output process, steps 3 to 7 are executed synchronously and repeatedly to calculate the updated partition energy compensation coefficient.

[0176] Step 85: Determine whether the thermal damage degree characterization values ​​of each control zone after the update in step 71 have all reached the preset target thermal damage degree threshold, and whether the energy accumulation state characterization values ​​of each control zone after the update in step 71 have all reached the corresponding target accumulation energy value.

[0177] Step 86: If the result of step 85 is that the condition is not met, return to step 83 to continue performing differential adjustment;

[0178] Step 87: If the judgment result of step 85 is that the condition is met, then it is determined that the condition for full-domain inactivation has been met, and the high-frequency current generator is controlled to terminate the energy output.

[0179] In this embodiment of the invention, the specific implementation process of step 81 is as follows:

[0180] From the output of step 75, extract the "spatial range identifier" (such as control area ID, boundary coordinate range) and the corresponding "zone energy compensation coefficient" (normalized 0-1 value) for all independent energy control zones. Organize these two types of data according to the "one-to-one correspondence" principle, for example, "Control area 1 (ID: 001) corresponds to a compensation coefficient of 0.93," "Control area 2 (ID: 002) corresponds to a compensation coefficient of 0.45," etc. Use a structured table format, with table fields including "Control area ID," "spatial boundary coordinates (Xmin, Xmax, Ymin, Ymax)," "zone energy compensation coefficient," and "data generation timestamp." After data entry, perform a completeness check on the table: check for any missing control areas (compare with the total number of control areas output in step 65), and check for any compensation coefficients exceeding the 0-1 range (if any, backtrack to step 75 to correct), ensuring that each control area has a unique corresponding compensation coefficient, providing an accurate mapping basis for subsequent instruction parsing.

[0181] The specific implementation process of step 82 above is as follows:

[0182] The conversion rules for preset compensation coefficients and increment ratios are as follows: When the compensation coefficient is 0, the increment ratio is 0% (no additional output required); when the compensation coefficient is 1, the increment ratio is 50% (maximum allowable increment, to avoid excessive energy output); intermediate coefficients are converted according to a linear relationship, i.e., "increment ratio = compensation coefficient × 50%". For example, a compensation coefficient of 0.93 corresponds to an increment ratio of 0.93 × 50% ≈ 46.5%, and a compensation coefficient of 0.45 corresponds to an increment ratio of 0.45 × 50% = 22.5%. For each control zone, based on the compensation coefficient in the mapping table, the required high-frequency current output increment ratio is calculated according to the above rules (the increment object includes voltage amplitude or duration, with priority given to adjusting the voltage amplitude; for example, if the initial voltage amplitude is 1400 volts, an increment of 46.5% would be adjusted to 1400 × (1 + 46.5%) ≈ 2051 volts).

[0183] The electrode control instruction set must include the following information: control area ID, corresponding electrode array number (electrode arrays are bound to control areas according to spatial partitions), target voltage amplitude (initial value + increment), target duration (if adjustable), and instruction execution timestamp. For example, "Control area ID: 001, Electrode number: E01, Target voltage: 2051 volts, Duration: 1.3 seconds (keeping initial), Execution time: t + 0.05ms".

[0184] Perform a validity check on the instruction set: check whether the target voltage is within the device output range (e.g., within the 1600-2000 volt range), whether the electrode number matches the control area, and ensure that the instruction can be recognized and executed by the high-frequency current generator. Finally, output a formatted instruction set (e.g., JSON format).

[0185] The specific implementation process of step 83 above is as follows:

[0186] The electrode control command set generated in step 82 is sent to the control system of the high-frequency current generator via a data interface. The control system parses the control zone ID, electrode number, and target parameters in the command. For each control zone corresponding to the electrode array, the control system adjusts the output parameters of its drive module to increase the voltage amplitude (or duration) of the high-frequency current according to the calculated increment ratio. For example, the electrode E01 corresponding to control zone 001 originally outputs 1400 volts, which is adjusted to 2051 volts by an increment of 46.5%, while keeping the frequency unchanged at 80kHz. During execution, the actual output parameters (voltage, current, duration) are monitored in real time by the built-in sensor of the generator and compared with the target value of the command. If the deviation exceeds 5% (preset allowable error), fine adjustment is immediately triggered (such as increasing the drive signal when the voltage is insufficient) to ensure that the current output of each control zone accurately matches the incremental demand.

[0187] The specific implementation process of step 84 above is as follows:

[0188] After the differentiated current output begins, a synchronous update cycle is set (e.g., once every 0.1 seconds, adjusted according to the thermal response speed). Within each cycle, the complete process from steps 3 to 7 is re-executed to generate new partition energy compensation coefficients. After each update, the new coefficients are compared with the coefficients of the previous cycle. If the change exceeds 10% (preset sensitivity threshold), it is marked as a "key control area" to provide priority reference for subsequent adjustments.

[0189] The specific implementation process of step 85 above is as follows:

[0190] Collect the updated data for each control zone after step 84, and verify the two conditions one by one. Check whether the "current thermal damage degree characterization value" (e.g., proportion) of each control zone is ≥ the preset target thermal damage degree threshold (e.g., 90%). For example, if the current proportion of control zone 001 is 92% ≥ 90%, the condition is deemed met; if the current proportion of control zone 002 is 88% < 90%, the condition is deemed not met. Check whether the "current energy accumulation state characterization value" (average value) of each control zone is ≥ the preset target accumulation energy value (e.g., 8℃·ms). For example, if the average value of control zone 001 is 8.2℃·ms ≥ 8℃·ms, the condition is met; if the average value of control zone 002 is 7.5℃·ms < 8℃·ms, the condition is not met. Only when both conditions of all control zones are met (i.e., 100% of control zones pass the verification) is the condition for full-domain inactivation determined to be met; if any control zone fails to meet the condition, the condition is deemed not met.

[0191] The specific implementation process of step 86 above is as follows:

[0192] If step 85 determines that the conditions are not met, the system automatically backtracks to step 83: based on the partition energy compensation coefficient updated in step 84, the electrode control instruction set is regenerated (the parsing process of step 82 is repeated, and the incremental ratio is adjusted according to the new coefficient), driving the high-frequency current generator to continue to implement differentiated current output for the non-compliant control area (focusing on increasing the energy increment of the non-compliant area); at the same time, the identifier of the non-compliant control area and the non-compliant items are recorded (such as "insufficient thermal damage in control area 002" and "insufficient energy accumulation in control area 003"). The compensation intensity of the corresponding area is increased in the new instruction set (such as appropriately increasing the upper limit of the incremental ratio to 60%) to accelerate the compliance process.

[0193] The specific implementation process of step 87 above is as follows:

[0194] If step 85 determines that the conditions are met (all control zones meet the standards), the system sends a "terminate output command" to the high-frequency current generator. The command includes the termination time and the gradual rate at which the voltage amplitude drops to 0 volts (e.g., 500 volts / ms to avoid sudden current stoppage impacting the equipment). After receiving the command, the generator gradually reduces the output voltage to 0 volts at the gradual rate while simultaneously cutting off the current supply to the electrode array. Subsequently, the system records the complete parameters of this weeding operation (energy output of each control zone, total time, time to meet the standard, etc.), generates a weeding completion log, and enters standby mode, waiting for the next round of weed location signals to trigger a new operation process. Through the above process, step 8 achieves dynamic closed-loop control of energy output, ensuring that weeds achieve full-area inactivation under the premise of precise energy consumption, which improves the reliability of weeding and avoids energy waste.

[0195] like Figure 2 As shown, embodiments of the present invention also provide a dynamic control system for the energy output of a high-frequency electric shock weeding device, comprising:

[0196] The acquisition module is used to identify the aboveground morphological features of target weeds through a visual sensor and generate location data including plant size and spatial distribution.

[0197] The calculation module is used to calculate the voltage amplitude, frequency, and duration parameters of the initial high-frequency current based on the plant size in the positioning data and according to a preset mapping relationship; execute the initial energy parameters and apply the high-frequency current to the weeds, while simultaneously acquiring the thermodynamic response signal of the weed tissue in real time;

[0198] The calibration module is used to calibrate multiple high thermal response focal points within the energy action zone based on the temperature rise rate distribution in the thermodynamic response signal; connect the high thermal response focal points to construct a dynamic action range boundary, and adjust the boundary morphology according to the real-time thermal diffusion trend;

[0199] The partitioning module is used to divide the boundary of the effective range into several independent energy control zones according to the thermodynamic gradient characteristics; based on the degree of thermal damage and energy accumulation state of each control zone, a partition energy compensation coefficient is generated; and the high-frequency current output is adjusted differentially according to the partition energy compensation coefficient until the thermodynamic response signal meets the global inactivation condition.

[0200] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for dynamic control of energy output in a high-frequency electric shock weeding device, characterized in that, The method includes: Step 1: Identify the aboveground morphological features of the target weeds using a visual sensor to generate location data including plant size and spatial distribution; Step 2: Based on the plant size in the positioning data, calculate the voltage amplitude, frequency and duration parameters of the initial high-frequency current according to the preset mapping relationship; Step 3: Execute the initial energy parameters and apply a high-frequency current to the weeds. Perform time series analysis on the infrared thermal radiation image sequence, extract the continuous temperature value of each spatial pixel during the current action, generate spatiotemporal temperature field data reflecting the real-time temperature change inside the weed tissue, and output the spatiotemporal temperature field data as a signal characterizing the thermodynamic response of the weed tissue. Step 4: Based on the temperature rise rate distribution in the thermodynamic response signal, calibrate multiple high thermal response focal points within the energy action zone; determine the spatial center coordinates of each effective high thermal response cluster, and calibrate these center coordinates as a high thermal response focal point within the energy action zone; output the set of coordinates of all calibrated high thermal response focal points. Step 5, connecting the high thermal response focal points to construct a dynamic boundary of the effective range, and adjusting the boundary shape according to the real-time heat diffusion trend, includes: Step 51, based on the coordinate set of all calibrated high thermal response focal points output in Step 4, performing a spatial topology connection operation to connect adjacent focal points that meet the preset distance threshold condition in pairs, forming an initial polygonal boundary network; Step 52, performing boundary smoothing processing on the initial polygonal boundary network to eliminate local sharp corners and generate a continuous closed initial effective range boundary outline; Step 53, based on the spatiotemporal temperature field data reflecting the real-time temperature change inside the weed tissue output in Step 3, calculating the heat diffusion rate field data at the current moment within the energy effective range; the heat diffusion rate field data reflects the heat... The spatial gradient characteristics propagating from the high-temperature zone to the low-temperature zone; Step 54, based on the thermal diffusion rate field data, predict the anisotropic thermal diffusion trend within a preset distance from the outer edge of the initial action range boundary contour; wherein, the region with a high thermal diffusion rate is predicted as the boundary expansion direction, and the region with a low thermal diffusion rate is predicted as the boundary maintenance or contraction direction; Step 55, based on the anisotropic thermal diffusion trend predicted in Step 54, perform real-time deformation adjustment on the initial action range boundary contour: adjust the boundary to extend outward by a preset deformation in the predicted expansion direction, and maintain or contract the boundary inward in the predicted maintenance or contraction direction; generate a dynamically updated action range boundary shape, and output the action range boundary shape data after dynamic adjustment at the current moment. Step 6: Divide the boundary of the effective range into several independent energy control zones according to the thermodynamic gradient characteristics, including: Step 61: Based on the dynamically adjusted boundary shape data of the effective range output in Step 55, extract the closed area enclosed by the boundary as the basic control range; Step 62: Based on the real-time updated spatiotemporal temperature field data output in Step 3, calculate the temperature change rate gradient value and heat accumulation dose value at each spatial location point within the basic control range; Step 63: Based on the temperature change rate gradient value and heat accumulation dose value calculated at each point in Step 62, generate a thermodynamic gradient distribution map covering the basic control range; Step 64: Perform isogradation line division on the thermodynamic gradient distribution map to form sub-regions with similar thermodynamic gradient characteristics; Step 65: Define each continuous sub-region with similar thermodynamic gradient characteristics obtained in Step 64 as an independent energy control zone, and output the spatial range identifier of all independent energy control zones; Step 7: Based on the thermal damage degree and energy accumulation state of each control zone, generate the partitioned energy compensation coefficient, including: Step 71: Obtain the thermal cumulative dose value of all points in each independent energy control zone at the current time; calculate the average thermal cumulative dose of all points in the control zone as the current energy accumulation state characterization value of the control zone; obtain the temperature value of all points in the control zone at the current time; count the proportion of pixels in the control zone whose temperature value reaches the tissue inactivation critical temperature threshold as the current thermal damage degree characterization value of the control zone; Step 72: Compare the current energy accumulation state characterization value with the preset target cumulative energy value of the control zone and calculate the energy accumulation difference; Step 73: Compare the current thermal damage degree characterization value with the preset target thermal damage degree threshold of the control zone and calculate the thermal damage degree difference; Step 74: Calculate the energy compensation demand intensity of the control zone based on the energy accumulation difference and the thermal damage degree difference; Step 75: Normalize the energy compensation demand intensity to generate the partitioned energy compensation coefficient of the control zone; generate the partitioned energy compensation coefficient for each independent energy control zone; Step 8, adjust the high-frequency current output differentially according to the partitioned energy compensation coefficient until the thermodynamic response signal meets the global inactivation condition, including: Step 81, based on the partitioned energy compensation coefficient of each independent energy control zone generated in Step 75, establish a mapping relationship table between the spatial identifier of the control zone and the compensation coefficient; Step 82, according to the mapping relationship table, analyze the required high-frequency current output increment ratio of each independent energy control zone, and generate the corresponding electrode control command set; Step 83, execute the electrode control command set to drive the high-frequency current generator to implement differentiated current output for each independent energy control zone divided in Step 6; Step 84, in the differential... During the differential current output process, steps 3 to 7 are executed synchronously and repeatedly to calculate the updated partition energy compensation coefficient; in step 85, it is determined whether the thermal damage degree characterization values ​​of each control zone updated in step 71 have reached the preset target thermal damage degree threshold, and whether the energy accumulation state characterization values ​​of each control zone updated in step 71 have reached the corresponding target accumulated energy value; in step 86, if the judgment result of step 85 is that the conditions are not met, the process returns to step 83 to continue executing differential adjustment; in step 87, if the judgment result of step 85 is that the conditions are met, the process determines that the full-domain inactivation condition has been reached, and controls the high-frequency current generator to terminate energy output.

2. The method for dynamic control of energy output of the high-frequency electric shock weeding device according to claim 1, characterized in that, Step 1 includes: Step 11: Simultaneously capture left-view and right-view images of the target area using a calibrated binocular vision sensor; Step 12: Perform vegetation spectral response enhancement processing on the left and right view images respectively, segment the weed pixel region and generate a binary contour mask; Step 13: Extract the feature matching point set based on the binary contour mask, and generate a depth mapping map by calculating the binocular disparity; Step 14: Convert the depth mapping map into a three-dimensional spatial point cloud model and fit it to the ground reference plane; Step 15: Calculate the vertical distance from the highest point of each connected domain in the point cloud model to the ground reference plane as the plant height, and the minimum circumcircle diameter of the point cloud projected onto the horizontal plane as the canopy diameter. Step 16: Using the coordinates of the centroid at the bottom of the point cloud connected domain as the plant position reference, output structured positioning data including three-dimensional coordinates, plant height, and canopy diameter.

3. The method for dynamic control of energy output of the high-frequency electric shock weeding device according to claim 2, characterized in that, Step 2 includes: Step 21: Based on the plant height value in the positioning data, refer to the preset mapping table between plant height range and voltage amplitude reference level to determine the corresponding voltage amplitude reference level; wherein, the mapping table is set so that different voltage reference levels correspond to different threshold ranges when the plant height value falls into different threshold ranges, and the higher the plant height value, the higher the voltage reference level. Step 22: Based on the canopy diameter value in the positioning data, calculate the initial action duration according to a preset positive correlation mapping relationship between the canopy diameter value and the initial action duration; wherein, the positive correlation mapping relationship is set such that the larger the canopy diameter value, the longer the calculated initial action duration; Step 23: Set a fixed high-frequency parameter value. This high-frequency parameter value is configured to be higher than the preset human safety frequency threshold and lower than the preset electromagnetic interference critical frequency threshold. Step 24: Take the voltage amplitude reference level determined in step 21 as the initial voltage amplitude, take the initial action duration calculated in step 22 as the initial action duration, and combine it with the fixed high-frequency parameter value set in step 23 to generate the voltage amplitude, frequency and action duration parameter combination of the initial high-frequency current.

4. The method for dynamic control of energy output of the high-frequency electric shock weeding device according to claim 3, characterized in that, Step 3 includes: Step 31: Based on the combination of voltage amplitude, frequency and duration parameters of the initial high-frequency current generated in step 24, control the high-frequency current generator to output a high-frequency current pulse sequence with corresponding voltage amplitude, frequency and planned duration. Step 32: Apply the high-frequency current pulse sequence to the target weed plants through a contact electrode array, wherein the spatial arrangement of the electrode array covers the spatial distribution range of the weeds indicated by the positioning data generated in step 1. Step 33: Throughout the application of the high-frequency current pulse sequence, an infrared thermal imager arranged in or near the gap of the electrode array is used to continuously capture a sequence of infrared thermal radiation images of the target weed plants and the surrounding area at a sampling rate higher than the frequency of the high-frequency current pulse.

5. The method for dynamic control of energy output of the high-frequency electric shock weeding device according to claim 4, characterized in that, Step 4: Based on the temperature rise rate distribution in the thermodynamic response signal, calibrate multiple high thermal response focal points within the energy action region, including: Step 41: Based on the spatiotemporal temperature field data output in step 35, calculate the temperature change per unit time of each spatial pixel within the energy action zone within a preset time window, and generate the corresponding temperature rise rate field data. Step 42: Perform spatial domain analysis on the temperature rise rate field data to identify all pixels whose temperature rise rate values ​​exceed a preset first rate threshold, forming a candidate high thermal response point set. Step 43: Perform spatial clustering on the candidate high thermal response point set, and aggregate the candidate points that meet the preset spatial proximity conditions into several independent high thermal response clusters. Step 44: Calculate the average temperature rise rate of all points in each high thermal response cluster, and select clusters whose average value exceeds a preset second rate threshold as effective high thermal response clusters.

6. A dynamic control system for the energy output of a high-frequency electric shock weeding device, wherein the system implements the method as described in any one of claims 1 to 5, characterized in that, include: The acquisition module is used to identify the aboveground morphological features of target weeds through a visual sensor and generate location data including plant size and spatial distribution. The calculation module is used to calculate the voltage amplitude, frequency, and duration parameters of the initial high-frequency current based on the plant size in the positioning data and according to a preset mapping relationship. The initial energy parameters are executed and a high-frequency current is applied to the weeds, while the thermodynamic response signal of the weed tissue is collected in real time. The calibration module is used to calibrate multiple high thermal response focal points within the energy action region based on the temperature rise rate distribution in the thermodynamic response signal. Connect the high thermal response focal points to construct a dynamic action range boundary, and adjust the boundary shape according to the real-time thermal diffusion trend; A partitioning module is used to divide the boundary of the effective range into several independent energy control regions according to the thermodynamic gradient characteristics; Based on the degree of thermal damage and energy accumulation state of each control zone, a zone energy compensation coefficient is generated; the high-frequency current output is adjusted differentially according to the zone energy compensation coefficient until the thermodynamic response signal meets the global inactivation condition.

Images