Unmanned aerial vehicle intelligent spraying method and system based on LiDAR mapping and tree core path optimization, storage medium and computer equipment

Abstract

The invention discloses an unmanned aerial vehicle intelligent spraying method and system based on LiDAR mapping and tree core path optimization, a storage medium and a computer device.The spraying method comprises the steps that firstly, a high-precision orchard three-dimensional point cloud map is constructed based on LiDAR data, and then the spatial position and crown parameters of a single fruit tree are accurately extracted through a semantic segmentation algorithm; the path is optimized through an improved bat algorithm, smooth and efficient spraying path planning is achieved, and therefore precise hovering and targeted pesticide applying operation can be executed, the deposition effect of fog drops on the middle and lower layers of a tree crown is improved, and pollution to a non-target area is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural aviation plant protection, in particular to an unmanned aerial vehicle intelligent spraying method, system, storage medium and computer device based on LiDAR mapping and tree core path optimization. BACKGROUND

[0002] With the development of precision agriculture, unmanned aerial vehicles gradually show important advantages in pesticide application in orchard environment, with characteristics of high operation efficiency, wide coverage, flexible operation, etc. However, the traditional agricultural unmanned aerial vehicles mostly adopt full coverage spraying mode, which has problems of large pesticide waste, low droplet utilization rate and serious pollution in non-target areas, especially in non-structured environment such as dense planting in orchard, overlapping of tree crown and complex terrain. Some current researches have tried to identify and segment fruit trees through image processing or deep learning methods to achieve target spraying control. However, these methods often rely on RGB images, are greatly affected by light changes and occlusions, and are difficult to maintain stability and accuracy in complex orchard environment. In addition, most path planning strategies do not consider the influence of heading angle change on flight stability and spraying efficiency, resulting in problems such as sharp turning of path, high energy consumption and large coverage error. SUMMARY

[0003] The first object of the present application is to overcome the deficiencies of the prior art and provide an unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization solves the problems of lack of three-dimensional spatial perception ability of unmanned aerial vehicles in non-structured orchard, inability to autonomously plan efficient paths and execute precise targeted pesticide application, and improves pesticide utilization rate and reduces environmental pollution risk.

[0004] The second object of the present application is to provide an unmanned aerial vehicle intelligent spraying system based on LiDAR mapping and tree core path optimization.

[0005] The third object of the present application is to provide a storage medium.

[0006] The fourth object of the present application is to provide a computer device.

[0007] The first object of the present application is achieved by the following technical solution: the unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization comprises the following steps:

[0008] S1, an autonomous unmanned aerial vehicle equipped with a laser radar is used to scan the orchard environment to obtain orchard data, and a high-precision three-dimensional point cloud map is constructed in real time based on the orchard data;

[0009] S2, preprocessing the three-dimensional point cloud map, and using a marker-controlled watershed algorithm to segment the tree crown semantics of the processed three-dimensional point cloud map to extract feature parameters of single fruit trees;

[0010] S3, based on the feature parameters, using an improved bat algorithm to optimize and plan the spraying path of the autonomous unmanned aerial vehicle to generate a three-dimensional flight trajectory;

[0011] S4, the autonomous unmanned aerial vehicle hovers precisely above the center of each fruit tree according to the three-dimensional flight trajectory to perform spot pesticide spraying.

[0012] Preferably, step S1 includes the following specific steps:

[0013] S11, the unmanned aerial vehicle scans the target orchard area under the support of autonomous flight control algorithm and multi-sensor fusion mechanism, the flight height is kept at 12 m and the speed is 1.5 m / s; point cloud data, IMU attitude information and GNSS geographic coordinates are collected and recorded in real time during the flight process;

[0014] S12, using simultaneous localization and mapping technology, the laser radar, IMU and GNSS data are fused and processed to realize the spatial registration and time sequence alignment of the point cloud, and then an incremental mapping method is used to generate a three-dimensional point cloud map of the orchard with a precision of centimeter level.

[0015] Preferably, step S2 includes the following steps:

[0016] S21, using a statistical outlier rejection algorithm and a cloth simulation filter to preprocess the original point cloud to obtain tree crown point cloud data;

[0017] S22, based on the tree crown point cloud data, a digital surface model and a digital elevation model are constructed by interpolation, and the difference between the two generates a tree crown height model:

[0018] ;

[0019] Wherein, CHM is the tree crown height model, DSM is the digital surface model, and DEM is the digital elevation model;

[0020] S23, using the detected tree crown vertex as the initial marker, using a marker-controlled watershed algorithm to segment the single tree crown; for the point cloud data of each fruit tree after segmentation, using an Alpha Shape algorithm to reconstruct its three-dimensional geometric shape and extract feature parameters including tree height and tree volume;

[0021] S24. Based on the flight track points and flight logs, establish the mapping relationship between the lidar coordinate system and the geographic coordinate system. Through the coordinate mapping transformation method, convert the coordinates of the tree crown center point extracted in the lidar coordinate system into standard latitude and longitude coordinates to determine the characteristic parameters of each fruit tree. These characteristic parameters include tree height, tree crown volume and geographic coordinates of the tree center.

[0022] Preferably, step S3 includes the following steps:

[0023] S31. Based on the characteristic parameters and the hovering height required for the drone spraying action, plan the drone's fixed-point spraying flight path.

[0024] S32. An improved bat algorithm is used to optimize the global path of the fixed-point spraying flight path, so as to improve the overall operational efficiency of path planning and the smoothness of the flight trajectory.

[0025] S33. In the path evolution process of the improved bat algorithm, a heading angle continuity constraint is introduced to limit the angle change range between adjacent flight segments, so as to achieve a smooth transition of the path between each spraying node, improve the continuity of UAV flight attitude control, and generate a three-dimensional flight trajectory.

[0026] Preferably, the improved bat algorithm is obtained by discretization improvement based on the bat algorithm, including the following steps:

[0027] A. Introduce a discrete update mechanism based on the existing bat algorithm:

[0028] The path is represented as an ordered sequence of waypoints. The path difference is defined as the number of waypoints between the current path and the globally optimal path. This path difference serves as a measure of the discrete search intensity and is used to dynamically adjust the path update magnitude.

[0029] Equation (1);

[0030] in, For the first In the nth iteration The update speed of each path; for The random disturbance factor within the interval; Representing a path with the optimal path The number of waypoints with differences; This is a round-down operation;

[0031] B. Then, multi-strategy neighborhood operations are introduced to update the path solution:

[0032] In each iteration, the algorithm adjusts based on speed. Different combinations of neighborhood strategies are selected to perturb and update the path, thereby improving the ability to escape local optima and enhancing global search performance.

[0033] C. A multi-objective total cost function integrating path length and heading change was constructed: Regarding path length, assuming the path consists of a sequence of points... Composed of, where the Euclidean distance between any two adjacent points is The total path length can then be expressed as Regarding course changes, the angle between three consecutive points is introduced as a measure of turning cost; any three points The corner formed The average value of the vector angle is calculated using the formula and used as the heading cost for the entire path.

[0034] Equation (2);

[0035] The path length and corner cost are normalized, and the following total cost function is constructed:

[0036] Equation (3);

[0037] in, and This is a weighting parameter used to balance the contribution ratio of path length to heading smoothness; and These are the maximum values ​​of path length and corner cost in the current population, respectively.

[0038] Preferably, the specific steps of using the improved bat algorithm to perform global path optimization on the fixed-point spray flight path in step S32 are as follows:

[0039] S321. Population initialization: Initialize the population containing multiple candidate paths. Each path corresponds to a complete set of tree core access orders to ensure that the initial solution covers the feasible solution space.

[0040] S322, Discrete velocity calculation: Calculate the difference between the current path and the global optimal path in terms of waypoint arrangement, and calculate the update velocity based on equation (1) to determine the search intensity and adaptively control the subsequent disturbance amplitude;

[0041] S323, Multi-strategy neighborhood perturbation: Perform multi-strategy neighborhood perturbation on the path solution according to the update speed to enhance the diversity and search depth of the path solution;

[0042] S324. Cost assessment and population update: After the disturbance is completed, the total cost of the fusion path length and heading smoothness is calculated based on equation (3), and the population is updated and superior individuals are retained based on whether the termination condition is met according to the cost.

[0043] S325. Termination Judgment: Through continuous iteration until the termination condition is met, the population gradually evolves and approaches the global optimum, and finally selects the path with the minimum total cost as the navigation scheme for fruit tree spraying operations.

[0044] Preferably, step S4 includes the following steps:

[0045] S41. Establish a communication channel through the MAVLink protocol to transmit the coordinates, hovering control parameters and spraying commands in the three-dimensional flight trajectory to the intelligent execution module of the UAV.

[0046] S42. The drone performs trajectory tracking and fixed-point control, enabling it to hover precisely above the heart of the fruit tree and activate the spraying device.

[0047] S43. Utilize the downwash airflow from the drone's rotor to create high-speed turbulence, disrupting the structure of tree canopy branches and leaves, enhancing droplet penetration and pesticide deposition uniformity, thereby completing the targeted pesticide spraying task.

[0048] The second objective of this invention is achieved through the following technical solution: a drone-based intelligent spraying system based on LiDAR mapping and tree center path optimization, used to achieve the first objective of a drone-based intelligent spraying method based on LiDAR mapping and tree center path optimization, comprising:

[0049] The drone platform module is used to perform flight and spraying tasks according to the planned path;

[0050] The perception module, located in the drone platform module, is used to collect environmental data;

[0051] The data processing module processes the collected environmental data to form a high-precision 3D point cloud map. Then, it processes the 3D point cloud map and uses the marker control watershed algorithm to perform tree canopy semantic segmentation on the processed 3D point cloud map to extract the feature parameters of individual fruit trees.

[0052] The airborne computing module, based on the characteristic parameters of a single fruit tree, uses an improved bat algorithm to optimize the spraying path of the autonomous drone, so as to calculate a three-dimensional flight trajectory including hovering point and spraying time.

[0053] The intelligent execution module receives the three-dimensional flight trajectory calculated by the airborne computing module and performs flight trajectory control and spraying tasks on the UAV platform module.

[0054] The third objective of this invention is achieved through the following technical solution: a storage medium, characterized in that it stores a program, which, when executed by a processor, implements the UAV intelligent spraying method based on LiDAR mapping and tree center path optimization as described in the first objective.

[0055] The fourth objective of this invention is achieved through the following technical solution: a computer device, including a processor and a memory for storing processor-executable programs, wherein when the processor executes the program stored in the memory, it implements the UAV intelligent spraying method based on LiDAR mapping and tree center path optimization as described in the first objective.

[0056] The present invention has the following advantages over the prior art:

[0057] 1. This invention constructs a high-precision 3D point cloud map of orchards based on LiDAR data. Combined with the Simultaneous Localization and Mapping (SLAM) algorithm, it completes the high-precision 3D reconstruction of unstructured orchard environments. It accurately extracts the spatial location and canopy parameters of individual fruit trees through semantic segmentation algorithm. Then, it uses an improved bat algorithm to optimize the path, realizing smooth and efficient spraying path planning. This enables precise hovering and targeted application of pesticides, improves the deposition effect of droplets in the middle and lower layers of the canopy, and reduces pollution in non-target areas.

[0058] 2. This invention addresses the problems of inaccurate identification, suboptimal path, and inaccurate spraying in existing drone spraying systems in unstructured orchard environments. It proposes a novel intelligent spot spraying system that integrates LiDAR mapping, tree center localization, and path optimization algorithms with angle constraints, which can significantly improve pesticide utilization and the level of intelligence in spraying operations.

[0059] 3. This invention introduces a marker-based watershed segmentation algorithm (MCWA), which enables accurate identification and contour extraction of individual fruit trees, and can extract key spatial parameters such as tree height and canopy volume, providing a precise sensing basis for path planning and spray control.

[0060] 4. This invention introduces the Improved Bat Algorithm (IBA) and combines it with a heading angle continuity constraint mechanism to effectively optimize path smoothness and flight energy consumption. While improving trajectory feasibility and operational efficiency, it significantly enhances the autonomous flight capability and path execution stability of UAVs in complex orchards.

[0061] 5. The system of this invention employs a closed-loop control system comprised of a perception module, a data processing module, an onboard computing module, and an intelligent execution module. This enables the UAV to automatically locate itself, autonomously plan its path, and precisely hover and spray in the orchard environment. The system possesses target recognition and trajectory tracking capabilities at the single-tree level, and can autonomously complete the fixed-point spraying process, significantly improving the level of automation, reducing reliance on manual labor and application errors, and demonstrating significant agricultural promotion and practical value. Attached Figure Description

[0062] Figure 1This is a flowchart of the UAV intelligent spraying method based on LiDAR mapping and tree center path optimization according to the present invention.

[0063] Figure 2 This is a flowchart of point cloud preprocessing, canopy semantic segmentation, and canopy parameter extraction according to the present invention.

[0064] Figure 3 This is a flowchart illustrating the path optimization using the improved bat algorithm in this invention.

[0065] Figure 4 This is a hardware module diagram of the UAV platform integrating LiDAR perception and intelligent path optimization according to the present invention.

[0066] Figure 5 This is a control flowchart of the UAV intelligent spraying system based on LiDAR mapping and tree center path optimization according to the present invention. Detailed Implementation

[0067] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0068] like Figure 1 As shown, this UAV intelligent spraying method based on LiDAR mapping and tree center path optimization includes the following steps:

[0069] S1. Use an autonomous drone equipped with LiDAR to scan the orchard environment to obtain orchard data, and build a high-precision three-dimensional point cloud map in real time based on this orchard data.

[0070] Step S1 includes the following specific steps:

[0071] S11. Supported by autonomous flight control algorithms and multi-sensor fusion mechanisms, the UAV performs a full-coverage scan of the target orchard area, maintaining a flight altitude of 12 m and a speed of 1.5 m / s. During the flight, it collects and records point cloud data, IMU attitude information, and GNSS geographic coordinates in real time.

[0072] S12. Using synchronous positioning and mapping technology, the data from lidar, IMU and GNSS are fused to achieve spatial registration and temporal alignment of point clouds. Then, an incremental mapping method is used to generate a three-dimensional point cloud map of the orchard with centimeter-level accuracy.

[0073] S2. The 3D point cloud map is preprocessed, and the label control watershed algorithm is used to perform tree canopy semantic segmentation on the processed 3D point cloud map in order to extract the feature parameters of individual fruit trees.

[0074] like Figure 2 As shown, step S2 includes the following steps:

[0075] S21. The original point cloud is preprocessed using a statistical outlier removal algorithm and a cloth simulation filter to obtain tree canopy point cloud data.

[0076] S22. Based on tree canopy point cloud data, a digital surface model and a digital elevation model are constructed through interpolation. The difference between the two is used to generate a tree canopy height model.

[0077] ;

[0078] Among them, CHM is the tree canopy height model, DSM is the digital surface model, and DEM is the digital elevation model;

[0079] S23. Using the detected tree crown vertices as initial labels, the label-controlled watershed algorithm is used to perform semantic segmentation of the single tree crown; for the point cloud data of each segmented fruit tree, the Alpha Shape algorithm is used to reconstruct its three-dimensional geometry and extract feature parameters including tree height and tree volume.

[0080] S24. Based on the flight track points and flight logs, establish the mapping relationship between the lidar coordinate system and the geographic coordinate system. Through the coordinate mapping transformation method, convert the coordinates of the tree crown center point extracted in the lidar coordinate system into standard latitude and longitude coordinates to determine the characteristic parameters of each fruit tree. These characteristic parameters include tree height, tree crown volume and geographic coordinates of the tree center.

[0081] Specifically, a system of linear equations is constructed using the planar radar coordinates (x, y) of the reference point and its corresponding geographic coordinates (latitude, longitude) to solve for the affine transformation parameters. Here, (xi, yi) represents the radar coordinates of the i-th reference point, and (lati, lngi) represents its corresponding geographic coordinates (i=1, 2, …, n). Then:

[0082]

[0083] Once the affine transformation parameters (a, b, c, d, e, f) are estimated, the geographic coordinates of any target point (x, y) can be calculated using the following affine transformation formula:

[0084]

[0085] The final output includes the tree height, canopy volume, and geographic coordinates of the tree core for each fruit tree, providing parameter support for path planning and precise spraying operations.

[0086] S3. Based on feature parameters, an improved bat algorithm is used to optimize and plan the spraying path of the autonomous UAV in order to generate a three-dimensional flight trajectory.

[0087] likeFigure 3 As shown, step S3 includes the following steps:

[0088] S31. Based on the characteristic parameters and the hovering height required for the drone spraying action, plan the drone's fixed-point spraying flight path.

[0089] S32. An improved bat algorithm is used to perform global path optimization on the fixed-point spray flight path to improve the overall operational efficiency of path planning and the smoothness of the flight trajectory; the improved bat algorithm is obtained by discretization improvement based on the bat algorithm, and includes the following steps:

[0090] A. Introduce a discrete update mechanism based on the existing bat algorithm:

[0091] The path is represented as an ordered sequence of waypoints. The path difference is defined as the number of waypoints between the current path and the globally optimal path. This path difference serves as a measure of the discrete search intensity and is used to dynamically adjust the path update magnitude.

[0092] Equation (1);

[0093] in, For the first In the nth iteration The update speed of each path; for The random disturbance factor within the interval; Representing a path with the optimal path The number of waypoints with differences; This is a round-down operation;

[0094] B. Then, multi-strategy neighborhood operations are introduced to update the path solution:

[0095] In each iteration, the algorithm adjusts based on speed. Different neighborhood strategy combinations are selected to perturb and update the path, thereby improving the ability to find local optima and enhancing global search performance. Among them, multi-strategy neighborhood operations include 2-opt operations (randomly selecting two waypoints in the path and reversing the path segment between them to eliminate possible path intersections), waypoint swapping (swapping the positions of any two waypoints to reconstruct the access order), and waypoint insertion (inserting a node into a new position in the path to adjust the local topology).

[0096] C. A multi-objective total cost function integrating path length and heading change was constructed: Regarding path length, assuming the path consists of a sequence of points... Composed of, where the Euclidean distance between any two adjacent points is The total path length can then be expressed as Regarding course changes, to avoid sharp turns in the path, thereby reducing flight energy consumption and operational instability, the angle between three consecutive points is introduced as a measure of turning cost; any three points The corner formed The average value of the vector angle is calculated using the formula and used as the heading cost for the entire path.

[0097] Equation (2);

[0098] To standardize the scale of different metrics and facilitate optimization, path length and corner cost are normalized, and the following total cost function is constructed:

[0099] Equation (3);

[0100] in, and This is a weighting parameter used to balance the contribution ratio of path length to heading smoothness; and These are the maximum values ​​of path length and corner cost in the current population, respectively.

[0101] The specific steps of using the improved bat algorithm to perform global path optimization on the fixed-point spray flight path in step S32 are as follows:

[0102] S321. Population initialization: Initialize the population containing multiple candidate paths. Each path corresponds to a complete set of tree core access orders to ensure that the initial solution covers the feasible solution space.

[0103] S322, Discrete velocity calculation: Calculate the difference between the current path and the global optimal path in terms of waypoint arrangement, and calculate the update velocity based on equation (1) to determine the search intensity and adaptively control the subsequent disturbance amplitude;

[0104] S323, Multi-strategy neighborhood perturbation: Perform multi-strategy neighborhood perturbation on the path solution according to the update speed to enhance the diversity and search depth of the path solution;

[0105] S324. Cost assessment and population update: After the disturbance is completed, the total cost of the fusion path length and heading smoothness is calculated based on equation (3), and the population is updated and superior individuals are retained based on whether the termination condition is met according to the cost.

[0106] S325. Termination Judgment: Through continuous iteration until the termination condition is met, the population gradually evolves and approaches the global optimum, and finally selects the path with the minimum total cost as the navigation scheme for fruit tree spraying operations.

[0107] S33. In the path evolution process of the improved bat algorithm, a heading angle continuity constraint is introduced to limit the angle change range between adjacent flight segments, so as to achieve a smooth transition of the path between each spraying node, improve the continuity of UAV flight attitude control, and generate a three-dimensional flight trajectory.

[0108] S4. The autonomous drone, based on a three-dimensional flight trajectory, hovers precisely above the heart of each fruit tree to perform targeted pesticide spraying.

[0109] Step S4 includes the following steps:

[0110] S41. Establish a communication channel through the MAVLink protocol to transmit the coordinates, hovering control parameters and spraying commands in the three-dimensional flight trajectory to the intelligent execution module of the UAV.

[0111] S42. The drone performs trajectory tracking and fixed-point control, enabling it to hover precisely above the heart of the fruit tree and activate the spraying device.

[0112] S43. Utilize the downwash airflow from the drone's rotor to create high-speed turbulence, disrupting the structure of tree canopy branches and leaves, enhancing droplet penetration and pesticide deposition uniformity, thereby completing the targeted pesticide spraying task.

[0113] like Figure 4 and Figure 5 As shown, the second objective of this invention is achieved through the following technical solution: a drone intelligent spraying system based on LiDAR mapping and tree center path optimization, used to achieve the first objective of a drone intelligent spraying method based on LiDAR mapping and tree center path optimization, comprising:

[0114] The drone platform module is used to execute flight and spraying tasks according to a planned path. Specifically, the drone platform module in this embodiment mainly consists of a drone, a power system, and a spraying device. The drone is built based on the E620P hexacopter architecture, equipped with six XRotor PRO X9-110 brushless motors and carbon fiber propellers. The propeller diameter is 34 inches and the pitch is 11 inches, providing stable and reliable flight power and payload capacity. The power system uses two 6S LiPo lithium batteries connected in series as the power source. The spraying device uses an integrated high-pressure brushless water pump with a maximum flow rate of 8 L / min and an operating pressure range of 0.3–0.5 MPa. It is also equipped with six fan-shaped atomizing nozzles, which can achieve efficient atomization and uniform spraying of the liquid, suitable for fixed-point spraying operations. At the same time, the drone platform module is equipped with a wireless communication module, which integrates the LQ-3 image and data transmission ground end and air end, realizing bidirectional real-time transmission of mission commands, image data, and attitude information. The control system in the drone uses the CUAV-V5+ flight control system, combined with the C-RTK dual-antenna positioning module and the FS-iA6B remote control module, to achieve precise trajectory control and remote operation.

[0115] The perception module, located on the UAV platform module, is used to collect environmental data. Specifically, the perception module consists of a Mid-360 hybrid solid-state LiDAR and a SYO11HD high-definition camera. The LiDAR has a maximum ranging capability of 40 m, an accuracy of ±2 cm, a 360° horizontal field of view and a 59° vertical field of view, and a scanning frequency of 10 Hz, enabling all-round perception and 3D mapping of the orchard environment. Simultaneously, this module uses the FAST-LIO2 algorithm to achieve high-precision fusion of multi-sensor data, completing the real-time construction of a 3D point cloud map of the orchard and the perception of the pose of individual fruit trees.

[0116] The data processing module processes the collected environmental data to construct a high-precision 3D point cloud map. It then processes the 3D point cloud map further and uses a marker-controlled watershed algorithm to perform canopy semantic segmentation on the processed map to extract the feature parameters of individual fruit trees. Specifically, it processes the collected 3D point cloud data to complete semantic segmentation, tree center localization, and canopy parameter extraction for individual fruit trees. Based on the extracted tree center geographic coordinates and structural features, it calls the Improved Bat Algorithm (IBA) to optimize and plan the UAV spraying path, generating a task path sequence that includes hovering position and operation timing.

[0117] The airborne computing module, based on the characteristic parameters of a single fruit tree, uses an improved bat algorithm to optimize the spraying path of the autonomous drone, so as to calculate a three-dimensional flight trajectory including hovering point and spraying time.

[0118] The intelligent execution module receives the 3D flight trajectory calculated by the onboard computing module and performs flight trajectory control and spraying tasks on the UAV platform module. The intelligent execution module establishes a communication connection with the flight control system via the MAVROS interface to receive and execute task path instructions. Its control process includes: the onboard computing module sending the optimized path information (including target waypoints and control parameters) to the UAV's flight control system via the MAVLink protocol; the flight control system providing real-time feedback of UAV attitude information for dynamic path correction; and once the flight platform reaches the target tree center, the onboard computing module controls the spraying module to start via a PWM signal, achieving precise, targeted atomized spraying directly above the tree canopy.

[0119] The third objective of this invention is achieved through the following technical solution: a storage medium, characterized in that it stores a program, which, when executed by a processor, implements the UAV intelligent spraying method based on LiDAR mapping and tree center path optimization as described in the first objective.

[0120] The fourth objective of this invention is achieved through the following technical solution: a computer device, including a processor and a memory for storing processor-executable programs, wherein when the processor executes the program stored in the memory, it implements the UAV intelligent spraying method based on LiDAR mapping and tree center path optimization as described in the first objective.

[0121] The above-described specific embodiments are preferred embodiments of the present invention and are not intended to limit the present invention. Any other changes or equivalent substitutions made without departing from the technical solution of the present invention are included within the protection scope of the present invention.

Claims

1. A method for intelligent spraying of unmanned aerial vehicle based on LiDAR mapping and tree core path optimization, characterized in that, The method comprises the following steps: S1, using an autonomous unmanned aerial vehicle equipped with a laser radar to scan the orchard environment to obtain orchard data, and constructing a high-precision three-dimensional point cloud map in real time based on the orchard data; S2, preprocessing the three-dimensional point cloud map, and using a marker-controlled watershed algorithm to segment the tree crown semantics of the processed three-dimensional point cloud map to extract the feature parameters of single fruit trees; S3, based on the feature parameters, using an improved bat algorithm to optimize and plan the spraying path of the autonomous unmanned aerial vehicle to generate a three-dimensional flight trajectory; S4, the autonomous unmanned aerial vehicle hovers precisely above the center of each fruit tree according to the three-dimensional flight trajectory, and performs spot pesticide spraying.

2. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization according to claim 1, characterized in that, Step S1 includes the following specific steps: S11, the unmanned aerial vehicle performs full coverage scanning on the target orchard area under the support of autonomous flight control algorithm and multi-sensor fusion mechanism, the flight height is kept at 12 m, and the speed is 1.5 m / s; point cloud data, IMU attitude information and GNSS geographic coordinates are collected and recorded in real time during the flight process; S12, using simultaneous localization and mapping technology, the laser radar, IMU and GNSS data are fused and processed to realize spatial registration and time sequence alignment of the point cloud, and then an incremental mapping method is used to generate a three-dimensional point cloud map of the orchard with a precision of centimeter level.

3. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization according to claim 1, characterized in that, Step S2 includes the following steps: S21, using a statistical outlier rejection algorithm and a cloth simulation filter to preprocess the original point cloud to obtain tree crown point cloud data; S22, based on the tree crown point cloud data, a digital surface model and a digital elevation model are constructed by interpolation, and the difference between the two generates a tree crown height model: ； Wherein, CHM is the tree crown height model, DSM is the digital surface model, and DEM is the digital elevation model; S23, using the detected tree crown vertex as the initial marker, a marker-controlled watershed algorithm is used for single tree crown semantic segmentation; for the point cloud data of each fruit tree after segmentation, an Alpha Shape algorithm is used to reconstruct its three-dimensional geometric shape, and feature parameters including tree height and tree volume are extracted; S24, the mapping relationship between the laser radar coordinate system and the geographic coordinate system is established according to the track point and the flight log, the tree crown center point coordinates extracted in the laser radar coordinate system are converted into standard latitude and longitude coordinates through coordinate mapping conversion method, to determine the feature parameters of each fruit tree, including tree height, tree crown volume and tree core geographic coordinates.

4. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization according to claim 1, characterized in that, Step S3 includes the following steps: S31, based on the feature parameters, the required hovering height of the unmanned aerial vehicle spraying, the flight path of the unmanned aerial vehicle for spot spraying is planned; S32, using an improved bat algorithm to globally optimize the spot spraying flight path to improve the overall operation efficiency of the path planning and the smoothness of the flight trajectory; S33, in the path evolution process of the improved bat algorithm, the heading angle continuity constraint is introduced to limit the angle change amplitude between adjacent flight segments, to realize the smooth transition of the path between the spraying nodes, and improve the continuity of the unmanned aerial vehicle flight attitude control, to generate a three-dimensional flight trajectory.

5. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization according to claim 4, characterized in that, The improved bat algorithm is obtained by discretization improvement based on the bat algorithm, and includes the following steps. A. A discrete updating mechanism is introduced based on the existing bat algorithm: The path is represented as a sequence of ordered waypoints The path difference is defined as the number of differences in the order of waypoints between the current path and the globally optimal path, which is used as a measure of the strength of the discrete search to dynamically adjust the update range of the path. , formula (1); wherein, is the th iteration of the th path update speed; is the random perturbation factor within the interval [0, 1]; denotes the number of waypoints of the path from the optimal path; is the floor operation; B. Then a multi-strategy neighborhood operation is introduced to update the path solution: In each iteration, the algorithm updates the path according to the velocity Different neighborhood strategy combinations are selected to perturb and update the path, so as to improve the ability to jump out of local optimum and enhance the global search performance. C. A multi-objective total cost function is constructed by fusing path length and heading change: in terms of path length, let the path be composed of point sequence , where the Euclidean distance between two adjacent points is , the total path length can be expressed as ; in terms of heading change, the included angle between three consecutive points is introduced as the steering cost measure; the turning angle composed of any three points is obtained by the vector included angle calculation formula, and its average value is taken as the heading cost of the whole path: , formula (2); The path length and the corner cost are normalized, and the following total cost function is constructed: , formula (3); wherein, and are weight parameters for balancing the contribution of path length and heading smoothness; and are the maximum values of path length and turn cost in the current population, respectively.

6. The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization according to claim 5, characterized in that, The specific steps of using the improved bat algorithm to globally optimize the fixed-point spraying flight path in step S32 are as follows: S321, population initialization: initialize the population containing multiple candidate paths, each path corresponds to a complete set of tree core access sequence, and ensure that the initial solution covers the feasible solution space; S322, discrete velocity calculation: calculate the difference between the current path and the global optimal path in the arrangement of waypoints, and calculate the update speed based on formula (1) to determine the search intensity and adaptively control the subsequent disturbance amplitude; S323, multi-strategy neighborhood disturbance: according to the update speed, the path solution is executed multi-strategy neighborhood disturbance to enhance the diversity and search depth of the path solution; S324, cost evaluation and population update: after the disturbance is completed, the total cost of fusing path length and heading smoothness is calculated based on formula (3), and whether the termination condition is met is judged according to the cost, and the population is updated and the excellent individuals are reserved. S325, termination determination: through continuous iteration until the termination condition is met, the population gradually evolves and tends to be global optimal, and finally the path with the smallest total cost is selected as the navigation scheme of the fruit tree spraying operation.

7. The LiDAR mapping and tree core path optimization based unmanned aerial vehicle intelligent spraying method according to claim 1, characterized in that, Step S4 includes the following steps: S41, establish a communication channel through MAVLink protocol, and transmit the coordinates in the three-dimensional flight trajectory, hovering control parameters and spraying instructions to the intelligent execution module of the unmanned aerial vehicle; S42, the unmanned aerial vehicle executes the flight path tracking and fixed-point control, so that the unmanned aerial vehicle accurately hovers above the fruit tree core and starts the spraying device; S43, use the rotor downwash airflow of the unmanned aerial vehicle to form a high-speed turbulent flow, which destroys the structure of the tree canopy branches and leaves, improves the penetration ability of the mist droplets and the uniformity of the pesticide deposition, and completes the fixed-point pesticide spraying task.

8. The unmanned aerial vehicle intelligent spraying system based on LiDAR mapping and tree core path optimization, characterized in that, The unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization for realizing any one of claims 1-5 comprises: An unmanned aerial vehicle platform module for executing flight tasks and spraying tasks according to a planned path; A perception module arranged on the unmanned aerial vehicle platform module for collecting environmental data; A data processing module for processing the collected environmental data to form a high-precision three-dimensional point cloud map, and then processing the three-dimensional point cloud map and using a marker-controlled watershed algorithm to perform tree canopy semantic segmentation on the processed three-dimensional point cloud map to extract feature parameters of a single fruit tree; An on-board computing module for optimizing and planning the spraying path of the autonomous unmanned aerial vehicle based on the feature parameters of the single fruit tree to calculate a three-dimensional flight trajectory containing hovering points and spraying times; An intelligent execution module for receiving the three-dimensional flight trajectory calculated by the on-board computing module and realizing flight trajectory control and spraying tasks for the unmanned aerial vehicle platform module.

9. Storage medium, characterized in that A program is stored, and the program is executed by a processor to realize the unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization in any one of claims 1-7.

10. Computer device, characterized in that A processor and a memory for storing a program executable by the processor are included, and the processor executes the program stored in the memory to realize the unmanned aerial vehicle intelligent spraying method based on LiDAR mapping and tree core path optimization in any one of claims 1-7.

Images