Method, system, device and product for automatic navigation of orchard sprayers

Abstract

The application discloses an orchard spraying machine automatic navigation method, system, device and product, relates to the automatic navigation field, and the method comprises the following steps: collecting orchard environment information and orchard spraying machine state information; when the orchard spraying machine travels between the fruit tree rows, the fruit tree target is identified based on the image between the fruit tree rows, the navigation center line is extracted, and visual navigation is carried out; when it is determined that the end of the fruit tree row has been reached through the image between the fruit tree rows, the visual navigation is switched to the dead reckoning mode; in the dead reckoning mode, the real-time position is calculated through dead reckoning, and the orchard spraying machine is controlled to travel from the end of the fruit tree row to the head turning point; when it is determined that the orchard spraying machine has reached the head turning point, the orchard spraying machine is controlled to turn at the head; after the orchard spraying machine completes the head turning, the visual navigation is restored, and the technical problem that the anti-interference ability is weak and the signal is unstable when a single sensor is used for navigation in the prior art can be solved.

Description

Technical Field

[0001] This application relates to the field of automatic navigation technology, and in particular to an automatic navigation method, system, equipment and product for an orchard sprayer. Background Technology

[0002] As modern agriculture develops towards automation and intelligence, automatic navigation of agricultural machinery has become a key technology for autonomous operation. Orchard sprayers are widely used in orchard plant protection operations and can adapt to orchards with small plots, steep slopes, and complex terrain with uneven surfaces.

[0003] However, orchard environments are complex and unpredictable. Satellite navigation signals are easily blocked by the canopy of fruit trees, leading to decreased positioning accuracy or signal interruption. Single-vision navigation performs poorly in low-light conditions, failing to reliably extract navigation paths. Furthermore, orchards have narrow row spacing and limited space at the edge of the field, with no navigation signal between the end of the row and the turning point at the edge, making path planning difficult and affecting the continuity of navigation operations. Existing navigation methods mostly use a single sensor, which suffers from weak anti-interference capabilities and unstable signals. Summary of the Invention

[0004] The purpose of this application is to provide an automatic navigation method, system, equipment and product for orchard sprayers, which can solve the technical problems of weak anti-interference ability and unstable signal when using a single sensor for navigation in the prior art.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an automatic navigation method for an orchard sprayer, comprising: Collect orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes location information, angular velocity information, and travel speed information; When the orchard sprayer travels between rows of fruit trees, the automatic navigation method identifies the fruit tree targets based on the image of the fruit tree rows, extracts the navigation centerline, and converts it into the lateral deviation and heading deviation of the orchard sprayer for visual navigation; wherein, the lateral deviation is the vertical distance between the navigation centerline and the current direction of the orchard sprayer on the horizontal plane, and the heading deviation is the angle between the navigation centerline and the current direction of the orchard sprayer; When the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode; in dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the driving speed information, and controls the orchard sprayer to travel from the end of the fruit tree row to the turning point at the beginning of the field. When the automatic navigation method determines that the orchard sprayer has reached the turning point at the edge of the field based on the real-time location, the automatic navigation method controls the orchard sprayer to turn at the edge of the field based on the location information; After the orchard sprayer completes the field turn, the automatic navigation method reverts to visual navigation; The process of the orchard sprayer moving between rows of fruit trees and turning at the edge of the field is repeated to achieve continuous automatic navigation operation between multiple rows in the orchard.

[0006] Optionally, the step of identifying fruit tree targets and extracting navigation centerlines based on the collected images of fruit tree rows using an improved target detection model specifically includes: The collected images of fruit tree rows were preprocessed, and the image resolution was standardized. An improved target detection model is used to detect targets in the preprocessed fruit tree row images, identify the root collar of the fruit tree, and use the midpoint of the root collar as the localization base point; the improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function; An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

[0007] Optionally, the step of classifying the positioning base points using an improved K-means clustering algorithm and removing noise points includes: When classifying the positioning base points, the slope of the left-side fruit tree row straight line fitted to each positioning base point is obtained. and the slope of the straight line of the fruit tree row on the right ; like If the preset first sign feature and slope absolute value threshold are not met, or If the preset second symbol feature and absolute value threshold are not met, the corresponding positioning base point will be judged as noise and removed. Wherein, the first symbolic feature is a positive value, and the second symbolic feature is a negative value.

[0008] Optionally, when the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode, specifically including: The automatic navigation method monitors the collected images between rows of fruit trees in real time. When it is determined that the number of straight-line fitting points of the fruit tree rows on both the left and right sides in a consecutive preset number of frames is less than a preset value, it is determined that the end of the fruit tree row has been reached, the visual navigation stops working, and the dead reckoning mode is switched.

[0009] Optionally, in the dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the driving speed information through dead reckoning, specifically including: When the visual navigation is working normally, dead reckoning is started simultaneously to record the initial position, initial heading deviation, speed and angular velocity of the orchard sprayer; The real-time heading deviation is obtained by integrating the angular velocity, and the real-time position coordinates are calculated by combining the travel speed and the real-time heading deviation.

[0010] Optionally, the driving speed is calculated in the following way: The travel speed is calculated based on the pixel-to-object ratio, the change in image pixels, and the travel time between the two fruit trees.

[0011] Secondly, this application provides an automatic navigation system for an orchard sprayer, comprising: a data acquisition unit for acquiring orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes position information, angular velocity information, and travel speed information; The control unit is configured to: identify fruit tree targets based on collected images of the fruit tree rows using an improved target detection model, extract the navigation centerline, and convert it into lateral and heading deviations for visual navigation when the orchard sprayer is traveling between rows of fruit trees; switch from visual navigation to dead reckoning mode when the images of the fruit tree rows indicate that the sprayer has reached the end of the row; calculate the real-time position based on the angular velocity and speed information using dead reckoning mode, and control the sprayer to travel from the end of the row to the turning point; control the sprayer to turn at the turning point based on the position information when the real-time position indicates that the sprayer has reached the turning point; and restore visual navigation after the sprayer completes the turning, repeating the above process to achieve continuous automatic navigation operations between multiple rows in the orchard. An execution unit is used to receive instructions from the control unit and drive the orchard sprayer to move.

[0012] Optionally, the control unit is further configured to: The collected images of fruit tree rows were preprocessed, and the image resolution was standardized. An improved target detection model is used to detect targets in the preprocessed image and identify the root collar of fruit trees, with the midpoint of the root collar as the localization base point. The improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function. An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the orchard sprayer automatic navigation method described in any one of the above.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the orchard sprayer automatic navigation method described above.

[0015] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the orchard sprayer automatic navigation method described above.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides an automatic navigation method, system, device, and product for orchard sprayers. It collects orchard environmental information and orchard sprayer status information. The orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes position, angular velocity, and speed. Through the fusion of multi-source information, it provides multi-source data support for navigation in complex orchard environments. When one information source is obstructed or interfered with, other information sources can continue to provide the necessary navigation information, avoiding the failure of a single sensor under signal obstruction or low light conditions. This solves the technical problems of weak anti-interference capability and unstable signal when using a single sensor for navigation in existing technologies. When the orchard sprayer travels between rows of fruit trees, the automatic navigation method identifies the fruit tree targets based on the images between the rows, extracts the navigation centerline, and converts it into lateral and heading deviations for visual navigation. Lateral deviation is the vertical distance between the navigation centerline and the current direction of the orchard sprayer on the horizontal plane, while heading deviation is the angle between the navigation centerline and the current direction of the orchard sprayer. These two navigation parameters allow for a precise description of the orchard sprayer's position relative to the navigation centerline, providing accurate navigation data for path tracking. When the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, it switches from visual navigation to dead reckoning mode. In dead reckoning mode, the automatic navigation method calculates the real-time position based on angular velocity and speed information and controls the orchard sprayer to travel from the end of the fruit tree row to the turning point at the edge of the field. Dead reckoning mode does not rely on external signals and uses angular velocity and speed information for autonomous positioning, solving the navigation problem of no navigation signal coverage between the end of the fruit tree row and the turning point at the edge of the field, ensuring the continuity of navigation operations. When the automatic navigation method determines that the orchard sprayer has reached the turning point at the edge of the field based on the real-time position, it controls the orchard sprayer to turn at the edge of the field based on the position information. High-precision positioning is achieved by restoring location information in open areas at the edge of the orchard. Combined with a segmented turning control method, this adapts to the limited space at the orchard edge, enabling precise turning. After the orchard sprayer completes its turn at the edge, the automatic navigation method reverts to visual navigation, repeatedly executing the process of the sprayer moving between rows of fruit trees and turning at the edge, achieving continuous automatic navigation across multiple rows within the orchard. Through the organic switching and cyclical execution of three modes—visual navigation, dead reckoning, and location-based navigation—full-process continuous automatic navigation is achieved, balancing stability, accuracy, and adaptability in the complex environment of the orchard. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an automatic navigation method for an orchard sprayer, provided as an embodiment of this application; Figure 2 A schematic diagram of a technical route provided in an embodiment of this application; Figure 3 A schematic diagram of a software system provided in an embodiment of this application; Figure 4 A hardware structure diagram of a satellite positioning and navigation system provided in an embodiment of this application; Figure 5 A software flowchart of a satellite navigation system provided in an embodiment of this application; Figure 6 This is a functional block diagram of image processing software provided in an embodiment of this application; Figure 7 This is a schematic diagram of the functional modules of an automatic navigation device for an orchard sprayer provided in one embodiment of this application; Figure 8 This is a general structural diagram of a tracked orchard sprayer navigation system provided in one embodiment of this application; Figure 9 A schematic diagram of a communication system provided in an embodiment of this application; Figure 10 This is a schematic diagram of a filtering and denoising process provided in an embodiment of this application; Figure 11 This is a schematic diagram of navigation parameter conversion provided in an embodiment of this application; Figure 12 This is a schematic diagram of a dead reckoning model provided in an embodiment of this application; Figure 13 This is a geometrical diagram of dead reckoning provided in an embodiment of this application; Figure 14 This is a schematic diagram of a turning path provided in an embodiment of this application; Figure 15 This is a schematic diagram of a navigation mode switching process provided in an embodiment of this application; Figure 16 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] In one exemplary embodiment, an automatic navigation method for an orchard sprayer is provided, the overall technical approach of which is as follows: Figure 2 As shown, the system comprises four parts: a GNSS navigation system, a vision-based navigation system, a dead reckoning system, and a fusion experiment. First, algorithm research and hardware / software design for the GNSS navigation system are conducted separately, implementing data parsing, path planning and tracking, and independent experiments. Simultaneously, a vision-based navigation model is built, completing data processing, feature extraction, model training and optimization, and inter-row straight-line tracking experiments. Then, the dead reckoning system is used to achieve the integration of the two navigation methods. Finally, a joint path tracking experiment based on vision-based inter-row straight-line tracking and GNSS ground-head turning is conducted. The software system architecture is as follows. Figure 3 As shown, the overall development work was carried out in three directions: visual navigation system, GNSS turning control at the orchard edge, and navigation switching from the end of the fruit tree row to the orchard edge. The visual navigation system uses cameras to collect images of fruit tree trunks, uses image processing algorithms to identify the trunks, calculates the navigation path and control parameters, and realizes the sprayer's movement control between fruit tree rows. The orchard edge GNSS turning control relies on GNSS to obtain the sprayer's positioning information, and the controller plans the turning path to realize the sprayer's turning action at the orchard edge. From the end of the fruit tree row to the orchard edge, the inertial navigation system is used to obtain the equipment's motion information to guide the movement, and switches to GNSS navigation mode when approaching the edge, thus completing the software control of the entire sprayer operation process.

[0022] The specific process of the embodiments of this application is as follows: Figure 1 As shown. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, it includes steps 101 to 106. Wherein: Step 101: Collect orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes location information, angular velocity information, and travel speed information.

[0023] Step 102: When the orchard sprayer travels between rows of fruit trees, the automatic navigation method identifies the fruit tree targets based on the image between the fruit tree rows, extracts the navigation centerline, and converts it into the lateral deviation and heading deviation of the orchard sprayer for visual navigation; wherein, the lateral deviation is the vertical distance between the navigation centerline and the current direction of the orchard sprayer on the horizontal plane, and the heading deviation is the angle between the navigation centerline and the current direction of the orchard sprayer.

[0024] As an optional implementation, the step of identifying fruit tree targets and extracting navigation centerlines based on the acquired images between rows of fruit trees using an improved target detection model specifically includes: The collected images of fruit tree rows were preprocessed, and the image resolution was standardized. An improved target detection model is used to detect targets in the preprocessed fruit tree row images, identify the root collar of the fruit tree, and use the midpoint of the root collar as the localization base point; the improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function; An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

[0025] In this embodiment, image processing software is used to implement functions such as image acquisition and preprocessing, target detection, and navigation path extraction. See the attached functional modules. Figure 6 Specifically, it includes four aspects: offline image analysis, real-time image processing, coordinate space transformation, and path tracking control. Among them, offline image analysis completes data acquisition, data augmentation, data annotation, and model training in sequence; real-time image processing realizes the calculation of visual information between rows in the orchard through data acquisition, feature extraction, centerline extraction, and row tail detection; coordinate space transformation is used to extract navigation parameters and study path tracking null value methods; and path tracking control carries out path planning and experimental verification.

[0026] The image acquisition and preprocessing process involves: real-time acquisition of video images between fruit tree rows using a monocular camera, followed by data augmentation processing including rotation, cropping, brightness adjustment, and exposure adjustment to expand the dataset size and improve the model's generalization ability. In this embodiment, a large number of fruit tree row image data were acquired in both a simulated orchard and an actual mango orchard environment. 1498 images were acquired in the simulated orchard, and 1520 images were acquired in the actual mango orchard. Through data augmentation processing such as rotation, cropping, brightness adjustment, and exposure adjustment, the sample size was increased to 5992 and 6080 images, respectively. All images were unified to a resolution of 640×640 pixels for subsequent training and inference of the improved object detection model.

[0027] In this embodiment, the CSPDarknet53 module of the YOLOv7-tiny model is replaced with a group convolution module. The group convolution module uses an improved ShuffleNet v1 module, which employs 1×1 group convolutions to change the feature channel dimension and combines them with 3×3 group convolutions to replace the original depthwise separable convolutions. The 1×1 group convolutions are used to adjust the information flow between channels, while the 3×3 group convolutions are used to extract spatial features. The combination of these two methods enhances the semantic interaction between channels, avoids the loss of semantic information, and effectively reduces the computational cost and parameter count of the model. A coordinated attention module is introduced. This module extracts coordinate information by pooling along the X and Y axes, performing average pooling on the horizontal and vertical features respectively to preserve spatial location information. After convolutional dimensionality reduction and activation processing, an attention weight map is generated, enhancing the feature representation of the target region. The original Leaky ReLU activation function is replaced with the SiLU activation function. The SiLU activation function has smooth non-linear characteristics, which can improve the convergence speed and final detection accuracy of the model during training.

[0028] In this embodiment, the improved object detection model is trained as follows: Labelme software is used to annotate the collected orchard images, with the annotation objects being the root collar and trunk of the fruit trees, generating XML format annotation files. The training set, validation set, and test set are divided into three groups at a ratio of 80%, 10%, and 10%, respectively. The initial learning rate is set to 0.001, and a cosine annealing strategy is used to decrease the learning rate. The Adam optimizer is used. The simulated orchard is used for 100 iterations of training, and the mango orchard for 30 iterations. After training, the improved object detection model achieves high detection speed, accuracy, recall, and mean average precision (mAP) in both the simulated orchard and the mango orchard.

[0029] In this embodiment, the navigation path extraction process is as follows: A trained improved target detection model is used to identify the root collars of fruit trees in the image, with the midpoint of the root collar serving as the positioning base point. An improved K-means clustering algorithm is used to classify the positioning base points, removing noise generated by other fruit tree rows, and dividing the positioning base points into two categories: left-side fruit tree rows and right-side fruit tree rows.

[0030] As an optional implementation, the step of classifying the positioning base points and removing noise points using an improved K-means clustering algorithm includes: When classifying the positioning base points, the slope of the left-side fruit tree row straight line fitted to each positioning base point is obtained. and the slope of the straight line of the fruit tree row on the right ; like If the preset first sign feature and slope absolute value threshold are not met, or If the preset second symbol feature and absolute value threshold are not met, the corresponding positioning base point will be judged as noise and removed. Wherein, the first symbolic feature is a positive value, and the second symbolic feature is a negative value.

[0031] In this embodiment of the application, the slope of the fitted straight line for the left row of fruit trees is assumed to be... The slope of the fitted straight line for the fruit tree row on the right is The absolute value threshold of the slope is ,but and , and The improved K-means clustering algorithm adds a slope constraint to the traditional K-means clustering algorithm. Since fruit tree rows in an orchard are usually arranged linearly, the slopes of the fitted lines for the left and right fruit tree rows have a clear sign characteristic: the slope of the left fruit tree row should be positive, and the slope of the right fruit tree row should be negative. This is addressed by setting a threshold for the absolute value of the slope. This allows for further constraints on the reasonable range of the slope, eliminating abnormal positioning base points caused by misidentification or occlusion. After filtering by constraints, positioning base points that conform to the distribution pattern of fruit tree rows are retained, and then K-means clustering is performed, which can effectively remove noise generated by other fruit tree rows and ensure the accuracy of the straight line fitting of the fruit tree rows. Figure 10 The process of filtering and denoising using an improved K-means clustering algorithm is demonstrated.

[0032] After clustering, the least squares method is used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows is calculated as the navigation center line between the fruit tree rows.

[0033] After extracting the navigation centerline, it needs to be converted into lateral deviation and heading deviation. A transformation model between the image coordinate system and the vehicle coordinate system is established. Combining the intrinsic parameters, installation height, and tilt angle of the monocular camera, the pixel distance between the navigation centerline in the fruit tree row image and the center vertical line of the fruit tree row image coordinate system is converted into the lateral deviation in the vehicle coordinate system. The angle between the navigation centerline and the current direction of the vehicle is used as the heading deviation.

[0034] In the embodiments of this application, such as Figure 11 As shown, the intrinsic parameters of a monocular camera include focal length parameters. , and principal point coordinates , This is obtained through camera calibration. The installation height of the monocular camera. and horizontal tilt angle The coordinates of the pre-aimed points (X_a, Y_a) in the fruit tree row image coordinate system are determined during installation. Using a coordinate system transformation formula, the coordinates of the pre-aimed points (X_a, Y_a) in the fruit tree row image coordinate system are converted to the corresponding point coordinates (X_c, Y_c, Z_c) in the vehicle body coordinate system. The specific transformation formula is as follows: ; In the formula, ( , ) represents the coordinates of the preview point in the image coordinate system between rows of fruit trees, where X_a represents the pixel coordinates of the preview point in the horizontal direction of the image, and Y_a represents the pixel coordinates of the preview point in the vertical direction of the image; The coordinates are the corresponding points in the vehicle coordinate system. The conversion process considers the influence of the monocular camera's mounting posture on the imaging geometry to ensure the accuracy of the navigation parameters. The pixel distance between the navigation centerline in the fruit tree row image and the vertical center line of the fruit tree row image coordinate system is converted into a lateral deviation in the vehicle coordinate system. The angle between the navigation centerline and the vehicle's current direction is used as the heading deviation to provide navigation parameters for subsequent path tracking. In this embodiment, the average lateral deviation of the navigation centerline extracted by visual navigation is 18.6 cm, and the average angle deviation is 3.8 degrees.

[0035] This implementation method reduces computational cost by replacing the original backbone network module with a group convolution module; enhances feature aggregation capabilities by introducing a coordinated attention module; and improves convergence speed by replacing the activation function with the SiLU activation function. Through these improvements, the improved target detection model can accurately identify the root collar of fruit trees. The improved K-means clustering algorithm effectively eliminates noise through slope constraints, ensuring the accuracy of navigation centerline extraction. A coordinate system transformation model converts the fruit tree row image information into navigation parameters in the vehicle coordinate system, providing accurate lateral and heading deviations for path tracking.

[0036] Step 103: When the automatic navigation method determines that it has reached the end of the fruit tree row through the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode; in dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the driving speed information through dead reckoning, and controls the orchard sprayer to travel from the end of the fruit tree row to the turning point at the beginning of the field.

[0037] As an optional implementation, when the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode, specifically including: The automatic navigation method monitors the collected images between rows of fruit trees in real time. When it is determined that the number of straight-line fitting points of the fruit tree rows on both the left and right sides in a consecutive preset number of frames is less than a preset value, it is determined that the end of the fruit tree row has been reached, the visual navigation stops working, and the dead reckoning mode is switched.

[0038] In this embodiment, the preset number is set to ten frames. As the orchard sprayer approaches the end of the fruit tree row, the number of fruit trees in the field of view of the monocular camera gradually decreases, and the number of root collar positioning reference points identified by the improved target detection model also decreases accordingly. When the number of straight-line fitting points on both sides of the fruit tree row in ten consecutive frames of images between fruit tree rows is less than one, it indicates that there are no effective tree targets in the field of view, and at this time it is determined that the end of the fruit tree row has been reached. Visual navigation stops working, and the navigation mode switches to dead reckoning mode.

[0039] As an optional implementation, in the dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the travel speed information through dead reckoning, specifically including: When the visual navigation is working normally, dead reckoning is started simultaneously to record the initial position, initial heading deviation, speed and angular velocity of the orchard sprayer; The real-time heading deviation is obtained by integrating the angular velocity, and the real-time position coordinates are calculated by combining the travel speed and the real-time heading deviation.

[0040] In this embodiment, the dead reckoning mode begins synchronous operation while the visual navigation is functioning normally, continuously recording relevant parameters. The initial position is provided by the positioning data from the last moment of the visual navigation, and the initial heading deviation is provided by the heading deviation data from the last moment of the visual navigation. Angular velocity can be acquired in real time by the inertial measurement unit.

[0041] Please see Figure 12 , Figure 12A global inertial rectangular coordinate system Oxy is established, with the x-axis pointing horizontally to the right and the y-axis pointing vertically upwards, to describe the absolute position of the sprayer in the orchard environment. The sprayer is simplified as a rectangular rigid body, with its center (center of mass) marked O1. The angle between the sprayer's forward axis (direction of motion) and the global vertical direction (parallel to the y-axis) is defined as the heading deviation. In the diagram, the lateral distance between the actual position of the sprayer and the desired navigation path is the position deviation P. e This model intuitively reflects two key navigation parameters of the sprayer during its journey: heading deviation. This describes the degree of directional deviation of the sprayer, and the positional deviation P. e This describes the lateral offset of the sprayer relative to the navigation centerline. These two parameters are the core inputs that the visual navigation mode needs to calculate in real time and use for path tracking control. The model also provides the initial state for dead reckoning (initial coordinates O1, initial value of θ, initial value of P). e Geometric description of the initial value.

[0042] In this embodiment of the application, the specific calculation formula for dead reckoning is as follows: ; ; ; In the formula, For the first k +1 sampling period at the end of the heading deviation of the orchard sprayer This is the initial heading deviation. For the first i The heading deviation of the orchard sprayer at the end of each sampling period and The first k At the end of +1 sampling period, the lateral and longitudinal coordinates of the orchard sprayer in the vehicle coordinate system are... and Let these be the initial horizontal and vertical coordinates of the orchard sprayer. For the first i Angular velocity measured by the gyroscope within each sampling period v For driving speed, T The system sampling period is i and k The sampling period number is... k It is an integer greater than or equal to 0.

[0043] Please see Figure 13This figure shows the key geometric parameters that need to be measured when dead reckoning begins (with the sprayer located at the end of the fruit tree row). Dead reckoning starts from the end of the fruit tree row (denoted as point M), and through periodic integral calculations, the position coordinates and heading deviation of the orchard sprayer are updated in real time to fit the travel path from the end of the fruit tree row to the turning point at the beginning of the field (denoted as point N). The travel time is determined by calculating the geometric parameters at the end of the fruit tree row and the travel speed. The formula for calculating the travel time is: In the formula, The travel time from the end of the fruit tree row to the turning point at the edge of the field. The vertical distance between the two trees at the end of the row and the turning point at the beginning of the field. Let v be the vertical distance between the two trees at the end of the row and the sprayer, and v be the travel speed. In the formula, The distance between the two trees at the end of the row. The camera's field of view angles were all obtained through actual measurements.

[0044] As an optional implementation, the driving speed is calculated in the following way: The travel speed is calculated based on the pixel-to-object ratio, the change in image pixels, and the travel time between the two fruit trees.

[0045] In this embodiment of the application, the specific formula for calculating the driving speed is as follows: In the formula, v For driving speed, This is the pixel-to-object ratio factor. and These represent the changes in pixel coordinates of the same root neck point of a fruit tree in the horizontal and vertical directions, respectively, in two consecutive frames of images between rows of fruit trees. The time interval between the two fruit trees.

[0046] In this embodiment, the spacing between fruit trees is obtained through actual measurement. A monocular camera captures consecutive frames of images between rows of fruit trees. By analyzing the pixel position changes of the same fruit tree in two adjacent frames, and combining the tree spacing and frame rate, the actual driving speed can be calculated. This method does not rely on external speed sensors and can estimate driving speed in real time using visual information.

[0047] This implementation method accurately determines the termination time of visual navigation through a row tail detection mechanism, avoiding navigation signal loss due to the absence of fruit trees as targets. The dead reckoning mode does not rely on external navigation signals; it uses an inertial measurement unit and visually estimated speed for autonomous positioning, solving the navigation problem of no navigation signal coverage between the end of the fruit tree row and the turning point at the edge of the field, ensuring the continuity of navigation operations. The visual speed estimation method eliminates the need for additional speed sensors, simplifying the system structure.

[0048] Step 104: When the automatic navigation method determines that the orchard sprayer has reached the turning point at the edge of the field based on the real-time location, the automatic navigation method controls the orchard sprayer to turn at the edge of the field based on the location information.

[0049] When the orchard sprayer reaches the turning point at the edge of the field using dead reckoning, it resumes signal reception, acquires the location information of the turning point, and simultaneously collects the field's characteristic parameters, including the tree row width and the length of the field edge. The tree row width is the distance between the trunks at the end of the rows on both sides.

[0050] In this embodiment, the turning point at the edge of the orchard is located outside the end of the fruit tree row. This area is unobstructed by the canopy of fruit trees, allowing the orchard sprayer's satellite positioning module to receive a stable satellite signal and restore centimeter-level positioning accuracy. The satellite positioning module obtains the accurate coordinates of the turning point, providing a benchmark for subsequent turning path planning. Simultaneously, characteristic parameters of the orchard edge are collected through the satellite positioning module or visual measurement: the tree row width is the distance between the end trunks of the trees on both sides of the current row, and the orchard edge length is the distance from the current fruit tree row to the entrance of the next fruit tree row.

[0051] In this embodiment, based on the turning radius constraint of the orchard sprayer, a segmented turning control method combining in-situ turning and straight-line driving is adopted for turning path planning. Because the orchard sprayer has a small turning radius, but is limited by the narrow space at the orchard's edge, a reasonable turning path planning is necessary. The segmented turning control method decomposes the entire orchard's turning process into multiple stages, including: leaving the current row from the end of the row of fruit trees, turning in place to adjust posture, driving straight to the entrance of the next row, and turning in place to align with the next row. The stages are sequentially connected to ensure a safe and smooth turn within a limited space.

[0052] In turning path planning, a neural network model can be used to match the optimal turning path. The neural network model takes tree row width and field head length as input samples, and the turning path and turning success rate as output samples. It is trained using previously collected point cloud data of orchard trees at the end of rows and manual turning trajectories. In this embodiment, a large amount of point cloud data of orchard trees at the end of rows is collected beforehand. Skilled operators control the orchard sprayer to turn at the field head, recording the complete driving trajectory and the success or failure of each turn. The neural network model is trained using tree row width and field head length as input features, and the corresponding turning path trajectory point sequence and turning success rate as output labels. After training, the neural network model can quickly match and output the optimal turning path planning scheme based on the current tree row width and field head length.

[0053] During steering control, based on the optimal steering path, speed control commands are generated for the drive mechanisms on both sides. By adjusting the speeds of these drive mechanisms, precise steering of the orchard sprayer is achieved. During steering, the actual driving trajectory is fed back in real time by an inertial measurement unit (IMU) and a tachometer, compared with the planned path, and the control commands are dynamically adjusted to ensure steering accuracy. The IMU provides real-time feedback on the orchard sprayer's actual heading angle, and the tachometer provides real-time feedback on the actual speeds of the drive mechanisms on both sides. The control module calculates the actual driving trajectory based on this, compares the actual trajectory with the planned path in real time, calculates the trajectory deviation, and dynamically adjusts the speed commands of the drive mechanisms on both sides using proportional-integral-derivative (PID) control algorithms, forming a closed-loop control to ensure the accuracy and stability of the steering process.

[0054] This implementation method utilizes a satellite positioning module to recover signals in open areas at the edge of the orchard, obtaining high-precision location information to provide an accurate benchmark for turning path planning. The segmented turning control method fully considers the structural characteristics of the orchard sprayer and the space limitations at the orchard edge, achieving safe turning within a limited space through multi-stage transitions. The neural network model, by learning from human operating experience, can quickly match the optimal turning path, improving the intelligence and adaptability of turning path planning. A closed-loop feedback control mechanism ensures the accuracy of trajectory tracking during the turning process.

[0055] Step 105: After the orchard sprayer completes the field turn, the automatic navigation method resumes visual navigation.

[0056] Step 106: Repeat the process of the orchard sprayer moving between rows of fruit trees and the process of the orchard sprayer turning at the edge of the field to achieve continuous automatic navigation operation between multiple rows in the orchard.

[0057] After the orchard sprayer completes its turn at the edge of the orchard, the automatic navigation method reverts to visual navigation. The satellite positioning module guides the sprayer to the starting position of the next row of fruit trees. Upon reaching the next row entrance, the monocular camera recaptures images of the fruit tree rows, visual navigation resumes, and the system switches back to visual navigation mode. The process of the sprayer moving between rows and making its turn at the edge of the orchard is then repeated. This involves repeatedly performing the aforementioned visual navigation, row tail detection and dead reckoning navigation, and turn path planning and control processes, achieving continuous automatic navigation across multiple rows within the orchard.

[0058] In the embodiments of this application, such as Figures 13-15 As shown, visual navigation completes the operations within the first row of fruit trees. Figure 14 In Phase L1, dead reckoning was completed for the MN section, and GNSS navigation completed two turns and straight-through maneuvers at the starting point. Figure 14 In the L2 and L3 stages, visual navigation completes the work within the second row of fruit trees. Figure 14 (Level 4).

[0059] Throughout the process, multi-sensor data fusion is used to adjust navigation parameters in real time, ensuring the stability and accuracy of the navigation system. In a mango orchard trial, between rows of fruit trees, the visual navigation mode provided high-precision path tracking. The maximum positional deviation for straight-line path tracking was 0.394 meters, the absolute mean was 0.186 meters, and the standard deviation was 0.138 meters. The orchard sprayer could operate stably and smoothly complete turning at the edge of the field. From the end of the row to the turning point at the edge of the field, the dead reckoning mode ensured continuous navigation in areas without signal; in the edge area, the location-based navigation mode provided precise steering control. The three navigation modes worked together, covering the complete path of the orchard sprayer's operation in the orchard, achieving truly fully automatic navigation.

[0060] This implementation method, through the organic switching and cyclical execution of three modes—visual navigation, dead reckoning, and location-based navigation—allows the orchard sprayer to adapt to various navigation scenarios in the complex environment of orchards. It overcomes the limitations of single-sensor navigation and achieves continuous, automated navigation throughout the entire process. The multi-sensor fusion strategy ensures the stability and accuracy of the navigation system under different signal conditions.

[0061] Steps 101 to 106 are performed by collecting orchard environmental information and orchard sprayer status information. The orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes position, angular velocity, and speed information. Through the fusion of multi-source information, multi-source data support is provided for navigation in complex orchard environments. When one information source is obstructed or interfered with, other information sources can continue to provide the necessary navigation information, avoiding the failure of a single sensor under signal obstruction or low light conditions. This solves the technical problems of weak anti-interference capability and unstable signal when using a single sensor for navigation in existing technologies. When the orchard sprayer travels between rows of fruit trees, the automatic navigation method identifies the fruit tree targets based on the images between the rows, extracts the navigation centerline, and converts it into lateral and heading deviations for visual navigation. Lateral deviation is the vertical distance between the navigation centerline and the current direction of the orchard sprayer on the horizontal plane, while heading deviation is the angle between the navigation centerline and the current direction of the orchard sprayer. These two navigation parameters allow for a precise description of the orchard sprayer's position relative to the navigation centerline, providing accurate navigation data for path tracking. When the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, it switches from visual navigation to dead reckoning mode. In dead reckoning mode, the automatic navigation method calculates the real-time position based on angular velocity and speed information and controls the orchard sprayer to travel from the end of the fruit tree row to the turning point at the edge of the field. Dead reckoning mode does not rely on external signals and uses angular velocity and speed information for autonomous positioning, solving the navigation problem of no navigation signal coverage between the end of the fruit tree row and the turning point at the edge of the field, ensuring the continuity of navigation operations. When the automatic navigation method determines that the orchard sprayer has reached the turning point at the edge of the field based on the real-time position, it controls the orchard sprayer to turn at the edge of the field based on the position information. High-precision positioning is achieved by restoring location information in open areas at the edge of the orchard. Combined with a segmented turning control method, this adapts to the limited space at the orchard edge, enabling precise turning. After the orchard sprayer completes its turn at the edge, the automatic navigation method reverts to visual navigation, repeatedly executing the process of the sprayer moving between rows of fruit trees and turning at the edge, achieving continuous automatic navigation across multiple rows within the orchard. Through the organic switching and cyclical execution of three modes—visual navigation, dead reckoning, and location-based navigation—full-process continuous automatic navigation is achieved, balancing stability, accuracy, and adaptability in the complex environment of the orchard.

[0062] Based on the same inventive concept, this application also provides an automatic navigation system for orchard sprayers to implement the above-described automatic navigation method for orchard sprayers. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the automatic navigation system for orchard sprayers provided below can be found in the limitations of the automatic navigation method for orchard sprayers described above, and will not be repeated here.

[0063] In one exemplary embodiment, such as Figure 7 As shown, an automatic navigation system for an orchard sprayer is provided, specifically including: The data acquisition unit is used to acquire orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes position information, angular velocity information, and travel speed information; The control unit is configured to: identify fruit tree targets based on collected images of the fruit tree rows using an improved target detection model, extract the navigation centerline, and convert it into lateral and heading deviations for visual navigation when the orchard sprayer is traveling between rows of fruit trees; switch from visual navigation to dead reckoning mode when the images of the fruit tree rows indicate that the sprayer has reached the end of the row; calculate the real-time position based on the angular velocity and speed information using dead reckoning mode, and control the sprayer to travel from the end of the row to the turning point; control the sprayer to turn at the turning point based on the position information when the real-time position indicates that the sprayer has reached the turning point; and restore visual navigation after the sprayer completes the turning, repeating the above process to achieve continuous automatic navigation operations between multiple rows in the orchard. An execution unit is used to receive instructions from the control unit and drive the orchard sprayer to move.

[0064] In this embodiment, the overall hardware structure of the orchard sprayer automatic navigation system is as follows: Figure 8 As shown, it mainly consists of three parts: information acquisition, control unit, and execution unit. Positioning and attitude information is acquired through a GNSS satellite antenna and an AHRS attitude measurement module, and image information is acquired using a camera and transmitted to a computer. The navigation controller receives positioning and attitude data, interacts with the computer, and allows for human-machine interaction via a display. Wireless remote control enables remote operation of the linear controller. The linear controller, as the core control component, works in conjunction with the navigation controller, computer, and engine, ultimately outputting commands to the execution unit. It controls the left and right tracks for walking and steering through electro-hydraulic proportional valves and plunger pumps, and controls the fan and water pump to complete spraying operations. Overall, the tracked orchard sprayer achieves multi-source information perception, intelligent decision-making, and precise execution.

[0065] In this embodiment, the above-mentioned automatic navigation system for orchard sprayers is built on the main body of the orchard sprayer. The automatic navigation system includes a sensing module, a control module, and an execution module. Specifically, the sensing module corresponds to the aforementioned data acquisition unit, the control module corresponds to the aforementioned control unit, and the execution module corresponds to the aforementioned execution unit. A communication system is used for data transmission between the modules.

[0066] The drive mechanism of the orchard sprayer can be wheeled or tracked, and it has three working modes: manual, remote control, and automatic, which can be switched freely according to operational needs. This embodiment uses a tracked model as an example. Its walking system consists of left and right tracks, and steering is achieved by controlling the speed difference between the two tracks. This tracked orchard sprayer adopts a dual-pump, dual-motor hydraulic drive structure. The hydraulic motors drive the track wheels to rotate, and the track walking speed is adjusted by changing the speed of the hydraulic motors.

[0067] The sensing module includes a monocular camera, a satellite positioning module, an inertial measurement unit, and a tachometer. The monocular camera, used to acquire image information between rows of fruit trees, is installed at the front end of the orchard sprayer, with its ground clearance determined based on the size of the sprayer and the spacing between the fruit tree rows. The satellite positioning module acquires the position and heading information of the orchard sprayer. In this embodiment, the satellite positioning module uses a Real-Time Kinematic Global Navigation Satellite System (RTK-GNSS) module, which can provide centimeter-level positioning accuracy. The specific hardware structure of the satellite positioning module in this embodiment is as follows: Figure 4 As shown in the diagram, this hardware architecture illustrates the hardware connection logic of a combined navigation control system based on the AT91S AM9263 processor. The integrated base station provides differential positioning information to the GNSS board via a differential module. The GNSS board receives satellite signals from master and slave satellite antennas and transmits the positioning data to the core processor via an RS232 interface. The AHRS attitude measurement module outputs attitude data to the processor via an RS232 interface. The motion controller interacts with the processor via a CAN bus to exchange control signals. An external 9~36V power supply is converted to 3.3V, 5V, and 12V by a power module to power the system. The processor also communicates with the vehicle-mounted display module via an RS232 interface to output and display positioning, attitude, and control data. The overall system constitutes a hardware system integrating positioning, attitude perception, motion control, and human-machine interaction. The corresponding satellite navigation system software workflow is as follows: Figure 5As shown, after the system starts up, it first completes the initialization of hardware, data fusion and navigation parameters. Then, it enables three serial port interrupts: AHRS, satellite signal and vehicle display. It then sequentially realizes attitude data calculation, positioning data packet acquisition and navigation control parameter acquisition, and summarizes multi-source information to the data processing and fusion module. Next, it acquires the motion status information of the equipment, and after the main control process unit calculates it, the linear controller outputs control commands to complete the closed-loop control of the entire navigation operation.

[0068] In this embodiment, the sensing module further includes an attitude sensor, which is an Attitude and Heading Reference System (AHRS) sensor used to acquire the vehicle's roll attitude. An Inertial Measurement Unit (IMU) is used for dead reckoning and can measure the angular velocity and acceleration of the orchard sprayer in real time. A tachometer is used to measure the rotational speed of the drive mechanism. In this embodiment, the tachometer measures the track wheel rotational speed; a contact-type tachometer is used, with a resolution of 0.1 revolutions per minute (rpm) and a measurement range covering 1 to 19999 rpm.

[0069] The control module employs an embedded controller to receive information from the perception module, perform data processing, path planning, and control command generation. In this embodiment, the embedded controller is equipped with the Python programming language, the PyTorch deep learning framework, and the CUDA parallel computing framework, enabling efficient execution of target detection models and navigation control algorithms.

[0070] The execution module includes an electro-hydraulic proportional valve and a hydraulic motor (engine), which receive commands from the control module to drive the tracks. The control module adjusts the opening of the electro-hydraulic proportional valve through a linear controller to control the speed of the hydraulic motor, thereby controlling the track's travel speed.

[0071] The following describes the process of experimentally identifying and determining the control model for the tracked drive system in this embodiment. A first-order inertial element is used to describe the control system transfer function. By measuring the track wheel speeds corresponding to different PWM duty cycles, the relationship between the PWM duty cycle and the track speed is fitted using the least squares method to achieve the conversion between the control quantity and the actual speed. Specifically, the orchard sprayer is suspended in the air, with the tracks suspended. The control module outputs pulse width modulation (PWM) signals with different duty cycles, ranging from 0% to 100% in 5% intervals. Measurements are taken for a preset duration (e.g., 5 seconds) for each duty cycle, and the corresponding track wheel speed is recorded. This process is repeated multiple times (e.g., 3 times) and the average value is taken. The experimental results show that the dead-zone duty cycle of the proportional valve for both forward and backward movement of the left and right tracks is approximately 20.5% to 20.6%, and the saturation duty cycle is approximately 60.4% to 65.4%. The relationship between the PWM duty cycle and the track speed is fitted using the least squares method to obtain the left and right track conversion relationships. The specific duty cycle for the left track is as follows: when the target speed of the left track is greater than or equal to 1155 rpm, the duty cycle is 65.4%; when the target speed of the left track is greater than 0 and less than 1155 rpm, the duty cycle is 0.03202 multiplied by the target speed of the left track plus 24.79; when the target speed of the left track is equal to 0, the duty cycle is 0; when the target speed of the left track is greater than -1155 rpm and less than 0, the duty cycle is -0.03693 multiplied by the target speed of the left track minus 20.51; when the target speed of the left track is less than or equal to -1155 rpm, the duty cycle is -63.2%. The specific duty cycle for the right track is as follows: when the target speed of the right track is greater than or equal to 1160 rpm, the duty cycle is 65.4%; when the target speed of the right track is greater than 0 and less than 1160 rpm, the duty cycle is 0.03862 multiplied by the target speed of the right track plus 20.63; when the target speed of the right track is equal to 0, the duty cycle is 0; when the target speed of the right track is greater than -1139 rpm and less than 0, the duty cycle is -0.03326 multiplied by the target speed of the right track minus 22.73; when the target speed of the right track is less than or equal to -1139 rpm, the duty cycle is -60.6%. By describing the corresponding relationship within different speed ranges using this piecewise linear function, the control module can accurately calculate the required PWM duty cycle based on the target speed, thereby achieving precise control of the track speed.

[0072] This implementation method, through experimental identification, obtains a precise control model for the orchard sprayer's walking system. This model accurately describes the correspondence between the PWM control signal and the actual rotational speed of the walking mechanism, providing a precise execution basis for subsequent path tracking control. The identification of dead zones and saturation zones ensures that the control signal operates within its effective range, preventing control failure.

[0073] In this embodiment of the application, the communication connection relationships between the modules are as follows: Figure 9 As shown: Figure 9 The demonstration showcased a multi-module communication architecture based on the CAN bus. The linear control system (such as the linear controller) and the GNSS navigation system achieve bidirectional data transmission through CANL and CANH differential signal lines. The engine (such as the hydraulic motor) ECU and the wireless remote control system (if wireless remote control is required) are all mounted on this CAN bus and can interact with the linear control system and the GNSS navigation system respectively, thereby realizing the real-time transmission and coordinated control of navigation commands, power control, and remote control signals.

[0074] As an optional implementation, the control unit is further configured to: The collected images of fruit tree rows were preprocessed, and the resolution of the fruit tree rows was unified. An improved target detection model is used to detect targets in the preprocessed fruit tree row images, identify the root collar of the fruit tree, and use the midpoint of the root collar as the localization base point; the improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function; An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

[0075] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 16 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores automatic navigation data for the orchard sprayer. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the automatic navigation method for the orchard sprayer.

[0076] Those skilled in the art will understand that Figure 16 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0077] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0078] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0079] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0080] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0081] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0082] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0083] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An automatic navigation method for an orchard sprayer, characterized in that, The automatic navigation method for the orchard sprayer includes: Collect orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes location information, angular velocity information, and travel speed information; When the orchard sprayer travels between rows of fruit trees, the automatic navigation method identifies the fruit tree targets based on the image of the fruit tree rows, extracts the navigation centerline, and converts it into the lateral deviation and heading deviation of the orchard sprayer for visual navigation; wherein, the lateral deviation is the vertical distance between the navigation centerline and the current direction of the orchard sprayer on the horizontal plane, and the heading deviation is the angle between the navigation centerline and the current direction of the orchard sprayer; When the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode; in dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the driving speed information, and controls the orchard sprayer to travel from the end of the fruit tree row to the turning point at the beginning of the field. When the automatic navigation method determines that the orchard sprayer has reached the turning point at the edge of the field based on the real-time location, the automatic navigation method controls the orchard sprayer to turn at the edge of the field based on the location information; After the orchard sprayer completes the field turn, the automatic navigation method reverts to visual navigation; The process of the orchard sprayer moving between rows of fruit trees and turning at the edge of the field is repeated to achieve continuous automatic navigation operation between multiple rows in the orchard.

2. The automatic navigation method for orchard sprayers according to claim 1, characterized in that, The process, based on the collected images between rows of fruit trees, involves identifying fruit tree targets using an improved target detection model and extracting the navigation centerline. Specifically, this includes: The collected images of fruit tree rows were preprocessed, and the image resolution was standardized. An improved target detection model is used to detect targets in the preprocessed fruit tree row images, identify the root collar of the fruit tree, and use the midpoint of the root collar as the localization base point; the improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function; An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

3. The automatic navigation method for an orchard sprayer according to claim 2, characterized in that, The step of classifying the positioning base points using an improved K-means clustering algorithm and removing noise points includes: When classifying the positioning base points, the slope of the left-side fruit tree row straight line fitted to each positioning base point is obtained. and the slope of the straight line of the fruit tree row on the right ; like If the preset first sign feature and absolute slope threshold are not met, or If the preset second symbol feature and absolute value threshold are not met, the corresponding positioning base point will be judged as noise and removed. Wherein, the first symbolic feature is a positive value, and the second symbolic feature is a negative value.

4. The automatic navigation method for an orchard sprayer according to claim 1, characterized in that, When the automatic navigation method determines that it has reached the end of the fruit tree row based on the image between the fruit tree rows, the automatic navigation method switches from visual navigation to dead reckoning mode, specifically including: The automatic navigation method monitors the collected images between rows of fruit trees in real time. When it is determined that the number of straight-line fitting points of the fruit tree rows on both the left and right sides in a consecutive preset number of frames is less than a preset value, it is determined that the end of the fruit tree row has been reached, the visual navigation stops working, and the dead reckoning mode is switched.

5. The automatic navigation method for an orchard sprayer according to claim 1, characterized in that, In the dead reckoning mode, the automatic navigation method calculates the real-time position based on the angular velocity information and the driving speed information through dead reckoning, specifically including: When the visual navigation is working normally, dead reckoning is started simultaneously to record the initial position, initial heading deviation, speed and angular velocity of the orchard sprayer; The real-time heading deviation is obtained by integrating the angular velocity, and the real-time position coordinates are calculated by combining the travel speed and the real-time heading deviation.

6. The automatic navigation method for an orchard sprayer according to claim 5, characterized in that, The driving speed is calculated in the following way: The travel speed is calculated based on the pixel-to-object ratio, the change in image pixels, and the travel time between the two fruit trees.

7. An automatic navigation system for an orchard sprayer, characterized in that, The orchard sprayer automatic navigation system includes: The data acquisition unit is used to acquire orchard environmental information and orchard sprayer status information; the orchard environmental information includes images between rows of fruit trees, and the orchard sprayer status information includes position information, angular velocity information, and travel speed information; The control unit is configured to: identify fruit tree targets based on collected images of the fruit tree rows using an improved target detection model, extract the navigation centerline, and convert it into lateral and heading deviations for visual navigation when the orchard sprayer is traveling between rows of fruit trees; switch from visual navigation to dead reckoning mode when the images of the fruit tree rows indicate that the sprayer has reached the end of the row; calculate the real-time position based on the angular velocity and speed information using dead reckoning mode, and control the sprayer to travel from the end of the row to the turning point; control the sprayer to turn at the turning point based on the position information when the real-time position indicates that the sprayer has reached the turning point; and restore visual navigation after the sprayer completes the turning, repeating the above process to achieve continuous automatic navigation operations between multiple rows in the orchard. An execution unit is used to receive instructions from the control unit and drive the orchard sprayer to move.

8. The automatic navigation system for orchard sprayers according to claim 7, characterized in that, The control unit is further used for: The collected images of fruit tree rows were preprocessed, and the image resolution was standardized. An improved target detection model is used to detect targets in the preprocessed image and identify the root collar of fruit trees, with the midpoint of the root collar as the localization base point. The improved target detection model replaces the original backbone network module with a group convolution module, introduces a coordinated attention module, and replaces the activation function. An improved K-means clustering algorithm was used to classify the positioning base points, remove noise points, and divide the positioning base points into two categories: left fruit tree row and right fruit tree row. The least squares method was used to fit the straight lines of the left and right fruit tree rows respectively, and the midline of the straight lines of the two fruit tree rows was calculated as the navigation center line between the fruit tree rows.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the orchard sprayer automatic navigation method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the automatic navigation method for an orchard sprayer as described in any one of claims 1-6.

Images