Orchard weeding robot dynamic path planning method and system based on visual perception

Abstract

The invention provides an orchard weeding robot dynamic path planning method and system based on visual perception, and the method comprises the steps: constructing an improved YOLO v8 network structure, carrying an OAK-D-PRO camera on a weeding robot to collect image data of fruit trees at two sides, and transmitting the image data to a navigation control module; the navigation control module recognizes trunks and obstacles between rows of an orchard by using the target detection model, the lower midpoint of an output trunk recognition frame is used as a tree row fitting reference point, a unilateral tree row line and a tree row center navigation line are fitted by using a least square method, a target navigation path is generated, and the execution mechanism controls the weeding robot to walk according to the target navigation path. In the walking process, the navigation control module dynamically adjusts the moving direction and the steering amplitude of the weeding robot in real time. According to the invention, the problem that the navigation path is greatly deviated due to environmental influence in the prior art is effectively solved, and efficient and accurate path planning of the orchard robot is realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of intelligent agriculture and robot technology, and particularly relates to a dynamic path planning method and system for orchard weeding robots based on visual perception. BACKGROUND

[0002] Under the background of the rapid development of intelligent agriculture, the automatic navigation technology of orchard weeding robots has become the key to improving the management efficiency of orchards. However, the existing orchard weeding robots face significant challenges in complex environments: the target recognition model relied on by traditional navigation path planning mainly relies on the combination of hand-designed feature extractors and classifiers in orchard navigation path planning, extracts visual features with distinguishing degrees in images, and then uses machine learning methods for classification or positioning. When identifying tree trunks in orchards using traditional methods, image preprocessing is generally performed first, then feature extraction is performed, HOG descriptor is used to extract tree trunk edge direction features, then a classifier is trained, SVM is used to train positive samples (tree trunks) and negative samples (non-tree trunks), then the detection stage is entered, different size windows are slid on the image, HOG features are extracted for each window and SVM is used to judge whether it is a tree trunk, and finally non-maximum suppression (NMS) is performed to merge overlapping detection boxes to obtain the final tree trunk position.

[0003] Traditional methods have some limitations, they rely on hand-designed features, have limited generalization ability, are not suitable for complex environments such as light changes and occlusions, have insufficient recognition accuracy for different shapes and sizes of tree trunks and obstacles in orchards, and are easily disturbed by environmental factors such as leaf occlusion and light changes; at the same time, tree row positioning often uses the midpoint of the tree root, which is easily blocked by soil and weeds, as the base point, and when fitting the tree row line with the conventional least squares method, it is easily affected by noise data, resulting in a large deviation between the actual navigation path and the preset path, which seriously restricts the work efficiency and automation level of the robot.

[0004] Therefore, there is an urgent need for a path planning technology that can adapt to the complex environment of orchards, improve target recognition accuracy, and improve the robustness of path fitting. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application provides a dynamic path planning method and system for orchard weeding robots based on visual perception, which effectively solves the problems of insufficient target recognition accuracy in traditional path planning methods, the recognition of tree trunks and obstacles by the model being easily disturbed by leaf occlusion and light changes, the fitting base point being easily blocked by soil or weeds, and the navigation path deviating from the preset trajectory due to the influence of noise data on the conventional least squares method.

[0006] The present application achieves the above technical purpose through the following technical means.

[0007] A visual perception-based orchard weeding robot dynamic path planning method, comprising the following processes:

[0008] Step 1: Collect photos under different weather conditions in different seasons to obtain an orchard road data set, and divide it into a training set, a validation set, and a test set according to the corresponding proportions;

[0009] Step 2: Construct an improved YOLO v8 network structure to obtain an improved YOLO v8 model: replace part of the C2f module in the backbone network with a ConvNeXt V2 module, and add a multi-scale expansion attention mechanism, i.e., the MSDA mechanism, after the SPPF layer in the backbone;

[0010] Step 3: Train, validate, and test the improved YOLO v8 model using the data set established in step 1; load the best weight file obtained after training into the improved YOLO v8 model to obtain a target detection model, and load the target detection model into a navigation control module;

[0011] Step 4: The weeding robot equipped with an OAK-D-PRO camera advances in the orchard, the OAK-D-PRO camera collects real-time image data of the trees on both sides of the orchard and transmits it to the navigation control module; the navigation control module uses the target detection model to identify the tree trunks and obstacles in the orchard in real time, ignores the trunk root area which is easily blocked, takes the midpoint of the lower edge of the trunk recognition box output by the target detection model as the tree row fitting reference point, and based on the fitting reference point, uses the least squares method to fit the single-side tree row line and the tree row center navigation line to generate a target navigation path;

[0012] Step 5: The executing mechanism controls the weeding robot to walk according to the target navigation path, in the walking process, the weeding robot collects visual data streams through the video acquisition device installed thereon and transmits them to the navigation control module, the navigation control module calculates the moving direction of the robot based on the visual data streams and the fitting algorithm, then compares and analyzes it with the target navigation path, calculates the deviation angle, and dynamically adjusts the moving direction and turning amplitude of the weeding robot.

[0013] Further, the specific process of step 4 is as follows:

[0014] First, after identifying the trees on both sides of the tree row through the target detection model, the coordinates of the opposite edges of the trunk recognition box are processed to obtain the midpoint coordinates of the lower edge of the trunk recognition box; then the coordinates of one side of the tree row are marked as , the fitting straight line equation is , the data set of the fitting straight line is , and the error sum of squares formula is:

[0015] (1)

[0016] wherein, , respectively represent the slope and intercept of the fitting straight line; , , represents the coordinate of the midpoint of the lower edge of each trunk recognition frame in the one-side tree row, represents the sum of squared errors, represents the y coordinate of the midpoint of the lower edge of the trunk recognition frame, represents the fitting straight line; represents the number of the trunk recognition frame in the one-side tree row; represents the number of the trunk recognition frame in the one-side tree row;

[0017] The partial derivatives of the parameters in the above formula (1) are respectively obtained as follows: ,

[0018] (2)

[0019] According to the above formula (1) and (2), the following is obtained:

[0020] (3)

[0021] The best estimation values of the parameters , in the formula (3) are obtained by solving the formula (3) as follows:

[0022] (4)

[0023] After the values of , are calculated, the expression of the one-side tree row line is obtained by combining the dataset of the midpoints of the lower edges of the trunk recognition frames as follows: , and the expression of the other-side tree row line is obtained as follows: ;

[0024] In the formula, represents the equation of the one-side fitting straight line, represents the slope of the equation of the one-side tree row line, represents the intercept of the equation of the one-side tree row line, represents the equation of the other-side fitting straight line, represents the slope of the equation of the other-side tree row line, represents the intercept of the equation of the other-side tree row line;

[0025] After the expressions of the two-side tree row lines are obtained, a plurality of points are selected as the reference datum for fitting at equal intervals on the one-side tree row line, and the coordinates of the one-side tree row reference points are P1 ​The longitudinal coordinates of the plurality of points are substituted into the expression of the tree row line on the other side respectively to obtain the corresponding coordinates P2 ;

[0026] Based on the plurality of reference points on the left and right sides, a plurality of reference points P of the navigation line located in the center of the tree row are calculated by coordinate mean value m wherein, , Subsequently, in combination with the central navigation line reference points, a straight line equation is fitted, and the formulas (1) to (4) are combined to solve the best straight line matching function of the navigation line reference points by minimizing the square error, to obtain the expression of the navigation line in the center of the orchard tree row wherein, represents the fitted navigation line equation, represents the slope of the fitted navigation line, represents the intercept of the fitted navigation line.

[0027] Further, in step 2, the ConvNeXt V2 module is a pure convolutional neural network architecture that introduces a sparse convolution mechanism to generate learning signals by randomly masking the original input. With the help of the full convolutional mask autoencoder FCMAE and the new global response normalization layer GRN, the inter-channel feature competition is enhanced.

[0028] Further, in step 2, the multi-scale expansion attention further enriches and improves the general representation ability of the YOLO v8 model by incorporating the self-attention mechanism. Its framework integrates three key links: multi-scale processing, attention calculation, and feature rearrangement, forming an optimized pattern recognition scheme. Specifically, the multi-scale expansion attention selectively focuses on the regions with higher attention scores in the input feature map for processing.

[0029] Further, in step 5, when the deviation angle exceeds the preset range, according to the deviation direction and angle value, the control parameters including the rotation direction and rotation angle are output to adjust the moving direction and turning amplitude of the weeding robot; if the deviation angle does not exceed the preset range or is completely coincident with the target navigation path, the weeding robot is controlled to maintain straight-line travel without adjustment.

[0030] A system for implementing the above-mentioned visual perception-based dynamic path planning method for orchard weeding robots, comprising

[0031] An OAK-D-PRO camera is deployed at the front end of the weeding robot to collect real-time image data of the trees on both sides of the orchard and transmit it to the navigation control module;

[0032] A navigation control module is mounted on a Jetson Xavier NX embedded platform, and is used for running a target detection model and a path planning algorithm, identifying tree trunks and obstacles between rows in an orchard, extracting a lower edge midpoint of a trunk identification frame as a tree row fitting positioning base point, fitting a single tree row line and a tree row center navigation line based on the positioning base point by using a least square method, generating a target navigation path, and detecting a path deviation in real time, and outputting a control parameter to an executing mechanism to adjust a moving direction of the weeding robot.

[0033] An executing mechanism controls a servo motor to drive a wheeled chassis of the weeding robot through a CAN bus to execute a path adjustment instruction.

[0034] The present application has the following beneficial effects:

[0035] By improving the YOLO v8 network architecture, replacing part of the c2f module with a ConvNeXt V2 module and adding MSDA, the target recognition accuracy is improved, the least square method of the lower edge midpoint of the identification frame positioning base point is combined to fit the tree row line and the navigation line, the deviation between the actual walking path and the target navigation path is determined and adjusted in real time, the problem of large deviation of the navigation path in the prior art due to environmental influence is solved, and efficient and accurate orchard robot path planning is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 An improved YOLO v8 model network architecture diagram;

[0037] Figure 2 A position deviation diagram of an actual walking route of the weeding robot and a target navigation path;

[0038] Figure 3 A path planning method flowchart;

[0039] Figure 4 A lower edge midpoint extraction diagram of a trunk identification frame. DETAILED DESCRIPTION

[0040] The present application will be further described below in conjunction with the accompanying drawings and specific embodiments, but the scope of protection of the present application is not limited thereto.

[0041] Referring to Figure 3 The orchard weeding robot dynamic path planning method based on visual perception, as shown in the drawings, comprises the following processes:

[0042] Step 1: Considering that the intensity of light and the variable shape of the detected object may affect the detection speed and accuracy, photo collection is performed under different weather conditions in different seasons, an orchard road data set is obtained, and the data set is divided into a training set, a verification set and a test set according to a ratio of 7:2:1.

[0043] Step 2: Construct an improved YOLO v8 network structure based on the YOLO v8 algorithm, and obtain the improved YOLO v8 model: as shown in the reference Figure 1 The ConvNeXt V2 module is used to replace part of the C2f module in the backbone network, and the multi-scale expansion attention (MSDA) mechanism is added after the SPPF layer in the backbone. The MSDA mechanism enhances the recognition robustness of the trunk shape and obstacles through multi-scale feature rearrangement and attention weight calculation. The above improvements not only enhance the recognition and pattern adaptation ability of the model, but also improve the feature extraction and further improve the efficiency. Therefore, the modified model can accurately identify different shapes and sizes of targets while maintaining high efficiency, thereby providing significant performance improvement for the segmentation task. Figure 1 In the above formula, ConvNX2 module refers to ConvNeXt V2 module, two-dimensional convolution is Conv2d, bounding box loss is Bbox Loss, classification loss is Cls Loss, connection layer is Concat, and up-sampling is Upsample.

[0044] The ConvNeXt V2 module is a pure convolutional neural network architecture that introduces a sparse convolution mechanism. It generates learning signals by randomly masking the original input. This design forces the model to predict the missing parts using the remaining context information, thereby significantly improving the model's learning and understanding ability of visual data. With the help of full convolutional mask autoencoder (FCMAE) and a new global response normalization layer (GRN), the architecture effectively enhances the feature competition between channels, making feature extraction more accurate and efficient.

[0045] The multi-scale expansion attention (MSDA) further enriches the model's general representation ability by incorporating a self-attention mechanism. Its framework integrates multi-scale processing, attention calculation, and feature rearrangement, forming an optimized pattern recognition solution. Specifically, it selectively focuses on the areas with higher attention scores in the input feature map for processing. This feature allows it to exhibit better adaptability and processing accuracy when dealing with different scales of image data,

[0046] Step 3: Train, validate, and test the improved YOLO v8 model using the dataset established in Step 1. Load the best weight file obtained after training into the improved YOLO v8 model to obtain a target detection model. Load the target detection model into the navigation control module.

[0047] Step 4: The weeding robot equipped with an OAK-D-PRO camera advances in the orchard. The OAK-D-PRO camera collects real-time image data of the fruit trees on both sides of the orchard and transmits it to the navigation control module. Refer to Figure 4As shown, the navigation control module uses the target detection model to identify the trunk and obstacles in the orchard in real time, ignores the trunk root area which is easily blocked, and takes the midpoint of the lower edge of the trunk recognition frame output by the target detection model as the tree row fitting reference point; based on the fitting reference point, the least square method is used to fit the unilateral tree row line and the tree row center navigation line, and the target navigation path, i.e. the travel route of the weeding robot in the working process, is generated;

[0048] The specific process is as follows:

[0049] First, the tree trunks of the two side tree rows are identified by the target detection model, and the coordinates of the midpoints of the lower edges of the trunk recognition frames are obtained by processing the coordinates of the opposite edges of the trunk recognition frames; preferably, the coordinates of the upper left corner of the trunk recognition frame are set as (x1, y1), and the coordinates of the lower right corner are set as (x2, y2), and the coordinates of the midpoint of the lower edge of the trunk recognition frame are set as ((x2+x1) / 2, y2). 1+ x2) / 2, y2);

[0050] After obtaining the coordinates of the midpoints of the lower edges of the trunk recognition frames, the coordinates of the one side tree row are denoted as , the fitting straight line equation is , the data set of the fitting straight line is , and the error sum of squares formula is:

[0051] (1)

[0052] wherein, 、 represent the slope and intercept of the fitting straight line, respectively; 、 、 represent the coordinates of the midpoints of the lower edges of the trunk recognition frames in the one side tree row, represents the error sum of squares, represents the y coordinate of the midpoint of the lower edge of the trunk recognition frame, represents the fitting straight line; represents the number of the trunk recognition frame in the one side tree row; represents the number of the trunk recognition frames in the one side tree row;

[0053] The partial derivatives of the parameters 、 in the above formula (1) are solved to obtain

[0054] (2)

[0055] According to the above formulas (1) and (2), we have:

[0056] (3)

[0057] Solving formula (3) gives the parameters , Best estimate of:

[0058] (4)

[0059] Calculated 、 After the value is obtained, combined with the midpoint data set below the trunk identification box, the expression of the tree line on one side can be obtained. Similarly, we can get the expression of the tree line on the other side: ;

[0060] In the above formula, represents the equation of the fitted line on one side (the left side in this embodiment), represents the slope of the tree line equation on one side, represents the intercept of the tree line equation on one side, represents the equation of the fitted line on the other side (the right side in this embodiment), represents the slope of the tree line equation on the other side, Represents the intercept of the tree line equation on the other side.

[0061] After obtaining the expression of the tree line on both sides, 8 points are equally spaced on one side of the tree line as the reference base for the center fitting. The corresponding coordinates of the reference points on one side of the tree line are P1 ,in =1, 2, ..., 8, substitute the ordinates of these 8 points into the tree line expression on the other side to obtain the corresponding coordinates P2 ,in =1, 2,…, 8.

[0062] Based on the 16 reference points on the left and right sides, 8 reference points P of the navigation line located in the center of the tree row are obtained by calculating the coordinate mean. m ,in, , , =1, 2, ..., 8, and then combine these 8 central navigation line reference points to fit the straight line equation, combine equations (1) to (4), and solve the best straight line matching function with the navigation line reference points by minimizing the square error, and obtain the navigation line expression in the center of the orchard tree row. ,in, represents the equation of the fitted navigation line, represents the slope of the fitted navigation line, Indicates the intercept of the fitted guide line.

[0063] Step 5: The actuator controls the weeding robot to walk according to the target navigation path, in the walking process, the weeding robot collects visual data stream through the video acquisition device installed thereon and transmits to the navigation control module, the navigation control module calculates the moving direction of the robot based on the visual data stream and the fitting algorithm, then compares and analyzes the target navigation path, calculates the deviation angle, and then dynamically adjusts the moving direction and steering amplitude of the weeding robot through the closed-loop control system, so that the orchard robot walks along the target navigation path; the specific process is as follows:

[0064] Referring to Figure 2 , the route where the initial moving direction of the weeding robot is located is the initial actual walking path of the weeding robot, and the deviation between the actual walking path of the weeding robot and the target navigation path has two cases, that is, deviating to one side of the target navigation path by more than a predetermined angle, and overlapping or deviating by an angle within a predetermined range.

[0065] For the above two cases, the specific way of determining the deviation in this embodiment is as follows:

[0066] The position information and moving direction of the weeding robot are continuously acquired, and the deviation between the actual walking path and the target navigation path is determined according to the position information and moving direction, and the deviation includes the deviation angle and the deviation direction.

[0067] The moving direction of the robot is corrected in real time based on the calculated deviation to ensure that it travels along the target navigation path. Specifically:

[0068] If the actual walking path deviates to the left or right by an angle exceeding the predetermined range as shown in Figure 2 , the control parameters including the rotation direction and the rotation angle are output according to the deviation direction and the angle value to adjust the moving direction and the steering amplitude of the weeding robot; if the deviation angle does not exceed the predetermined range, or is completely overlapped with the target navigation path, the weeding robot is controlled to keep straight driving without adjustment.

[0069] The dynamic path planning system of the orchard weeding robot based on visual perception provided by the application comprises:

[0070] The OAK-D-PRO camera is deployed at the front end of the weeding robot, the installation height is 1.2 m from the ground, the pitch angle is 15°, and is used to collect image data of the trees on both sides of the orchard in real time and transmit to the navigation control module;

[0071] The navigation control module is carried on a Jetson Xavier NX embedded platform, is used for running a target detection model and a path planning algorithm, recognizes orchard inter-row tree trunks and obstacles, extracts the lower edge midpoint of the trunk recognition frame as a tree row fitting positioning base point, fits a single-side tree row line and a tree row center navigation line based on the positioning base point by using a least square method, generates a target navigation path, and detects a path deviation in real time, and outputs a control parameter to an executing mechanism to adjust the moving direction of the weeding robot, so that the weeding robot walks along the target navigation path.

[0072] The executing mechanism controls the servo motor to drive the weeding robot wheel type chassis through the CAN bus, and the steering precision is ±0.5°, and the executing path adjustment instruction is executed.

[0073] When the target navigation path is calculated, the midpoints of each pair of opposite fruit trees on both sides are taken as references, and the midpoints are sequentially connected to form a center line, which is the moving path of the robot. Based on the above-mentioned target navigation path determination method, the reasonable selection of features is the key to realize accurate navigation when the weeding robot drives in the orchard environment. The orchard environment is often complex, and the roots of fruit trees may be blocked due to branching and overgrowth of weeds, making it difficult to identify the root features. The fruit trees themselves have obvious available features, that is, the trunk and the crown are separated obviously, and the trunk height is high and not easy to be blocked by weeds. In addition, the position deviation between the midpoint of the bottom of the trunk and the center of the root is small. Combining these advantages, the midpoint of the bottom of the trunk of the fruit tree is selected as the reference point for fitting the navigation path.

[0074] The embodiments are preferred embodiments of the present application, but the present application is not limited to the above-mentioned embodiments, and any obvious improvement, replacement or modification made by those skilled in the art without departing from the essential content of the present application belongs to the protection scope of the present application.

Claims

1. A dynamic path planning method for an orchard weeding robot based on visual perception, characterized in that: The process includes the following: Step 1: Collect photos under different weather conditions in different seasons to obtain an orchard road dataset, and divide it into training set, validation set, and test set according to the corresponding proportions; Step 2: Build an improved YOLO v8 network structure to obtain an improved YOLO v8 model: Use ConvNeXt V2 modules to replace some C2f modules in the backbone network, and add a multi-scale expansion attention mechanism, namely the MSDA mechanism, after the SPPF layer in the backbone. Step 3: Use the dataset created in step 1 to train, verify, and test the improved YOLO v8 model; load the optimal weight file obtained after training into the improved YOLO v8 model to obtain a target detection model, and load the target detection model into the navigation control module; Step 4: The weeding robot, equipped with an OAK-D-PRO camera, moves through the orchard. The OAK-D-PRO camera collects real-time image data of fruit trees on both sides of the orchard and transmits it to the navigation control module. The navigation control module uses the target detection model to identify tree trunks and obstacles between orchard rows in real time, ignoring the easily obscured root areas of the tree trunks. The navigation control module uses the lower midpoint of the tree trunk identification box output by the target detection model as the tree row fitting reference point. Based on the fitted reference point, the least squares method is used to fit the single-side tree row line and the tree row center navigation line to generate the target navigation path. Step 5: The actuator controls the weeding robot to walk according to the target navigation path. During the walking process, the weeding robot collects visual data streams through the video acquisition device installed on it and transmits them to the navigation control module. The navigation control module calculates the robot's movement direction based on the visual data stream and the fitting algorithm, and then compares and analyzes it with the target navigation path, calculates the deviation angle, and dynamically adjusts the movement direction and steering range of the weeding robot.

2. The method for dynamic path planning of an orchard weeding robot based on visual perception according to claim 1, characterized in that: The specific process of step 4 is as follows: First, after identifying the trees on both sides of the tree row through the target detection model, the coordinates of the opposite sides of the trunk identification frame are processed to obtain the coordinates of the lower midpoint of the trunk identification frame; then the coordinates of the tree row on one side are marked as , the equation of the fitting line is , the data set of the fitted straight line is , the error sum square formula is: （1） in, 、 They represent the slope and intercept of the fitted line respectively; 、 、 Indicates the coordinates of the bottom midpoint of each trunk identification box in a row of trees on one side. represents the sum of squared errors, Indicates the y coordinate of the bottom midpoint of the trunk identification box, represents the fitted straight line; Indicates the trunk identification frame number in a tree row on one side; Indicates the number of tree trunk identification boxes in a tree row on one side; For the parameters in formula (1) 、 Taking partial derivatives respectively, we get: （2） According to the above formulas (1) and (2), we can obtain: （3） Solve equation (3) to get the parameters 、 Best estimate of: （4） Calculated 、 After the value is obtained, the expression of the tree line on one side is obtained by combining the midpoint data set below the tree trunk identification box. Similarly, we can get the expression of the tree line on the other side: ; Where, represents the equation of the fitted line on one side, represents the slope of the tree line equation on one side, represents the intercept of the tree line equation on one side, represents the equation of the fitted line on the other side, represents the slope of the tree line equation on the other side, represents the intercept of the tree line equation on the other side; After obtaining the expression of the tree line on both sides, multiple points are equally spaced on one side of the tree line as the reference base for the center fitting. The corresponding coordinates of the reference points on one side of the tree line are P1 , substitute the vertical coordinates of these multiple points into the tree line expression on the other side to obtain the corresponding coordinates P2 ; Based on multiple reference points on the left and right sides, multiple reference points P of the navigation line located in the center of the tree row are obtained by calculating the coordinate mean. m ,in, , Then, combining these central navigation line reference points, fitting the straight line equation, combining equations (1) to (4), and solving the best straight line matching function with the navigation line reference points by minimizing the square error, the navigation line expression of the center of the orchard tree row is obtained. ,in, represents the equation of the fitted navigation line, represents the slope of the fitted navigation line, Indicates the intercept of the fitted guide line.

3. The method for dynamic path planning of an orchard weeding robot based on visual perception according to claim 1, characterized in that: In step 2, the ConvNeXt V2 module is a pure convolutional neural network architecture that introduces a sparse convolution mechanism to generate learning signals by randomly masking the original input. This architecture uses a fully convolutional masked autoencoder (FCMAE) and a new global response normalization layer (GRN) to enhance feature competition between channels.

4. The method for dynamic path planning of an orchard weeding robot based on visual perception according to claim 1, characterized in that: In step 2, the multi-scale dilated attention further enriches and improves the general representation capabilities of the YOLO v8 model by incorporating the self-attention mechanism. Its framework integrates three key links: multi-scale processing, attention calculation, and feature reordering to form an optimized pattern recognition solution. Specifically, the multi-scale dilated attention selectively focuses on areas with higher attention scores in the input feature map for processing.

5. The method for dynamic path planning of an orchard weeding robot based on visual perception according to claim 1, characterized in that: In step 5, when the deviation angle exceeds the preset range, control parameters including the rotation direction and rotation angle are output according to the deviation direction and angle value to adjust the moving direction and steering amplitude of the weeding robot; if the deviation angle does not exceed the preset range, or completely overlaps with the target navigation path, the weeding robot is controlled to maintain straight-line movement without adjustment.

6. A system for implementing the dynamic path planning method of an orchard weeding robot based on visual perception as described in claim 1, characterized in that: include The OAK-D-PRO camera is deployed at the front end of the weeding robot to collect real-time image data of fruit trees on both sides of the orchard and transmit it to the navigation control module; The navigation control module, powered by the Jetson Xavier NX embedded platform, runs target detection models and path planning algorithms to identify tree trunks and obstacles between orchard rows. It extracts the lower midpoint of the trunk identification box as the tree row fitting positioning base point. Based on this positioning base point, it uses the least squares method to fit the unilateral tree row line and the tree row center navigation line to generate the target navigation path. It also detects path deviations in real time and outputs control parameters to the actuator to adjust the weeding robot's movement direction. The actuator controls the servo motor to drive the wheeled chassis of the weeding robot through the CAN bus to execute the path adjustment instruction.

Images