Double-unmanned-aerial-vehicle cooperative intelligent pesticide spraying method for fruit trees

Abstract

The invention discloses an unmanned aerial vehicle double-unmanned aerial vehicle cooperative intelligent pesticide spraying method for fruit trees, and relates to the technical field of plant protection unmanned aerial vehicles, and the method comprises the steps: deploying a laser radar and a high-precision GPS on an unmanned aerial vehicle, collecting orchard three-dimensional point cloud data, and constructing an orchard three-dimensional point cloud data set; performing data preprocessing on data in the constructed orchard three-dimensional point cloud data set, cleaning the data and performing position correction; an orchard three-dimensional grid model is constructed, orchard areas are divided based on the height of fruit trees, and tasks are allocated to the unmanned aerial vehicle; planning a pesticide spraying path of the unmanned aerial vehicle to execute pesticide spraying operation, and monitoring pesticide sedimentation and spraying effects in real time; the comprehensive score YL of the orchard is obtained according to pesticide spraying data collected in real time, the comprehensive score YL and a preset pesticide spraying evaluation threshold value YZ are compared and analyzed, the current pesticide spraying condition of the orchard is evaluated, and supplementary operation is conducted; pesticide waste caused by height difference of fruit trees is effectively reduced, the coverage rate and uniformity of pesticide spraying are improved, and the working efficiency of the unmanned aerial vehicle is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of plant protection UAVs, and in particular to a method for intelligent spraying of fruit trees using dual UAVs in a coordinated manner. Background Art

[0002] In modern agriculture, precision agriculture is gradually gaining attention as an important means to improve agricultural production efficiency and sustainable development. Especially in the field of fruit tree planting, the use of drone technology for intelligent management has become an emerging trend. This field not only covers the application of drones, but also involves technologies such as image recognition, data analysis and environmental monitoring. Specifically, in the management of spraying pesticides for fruit trees, the dual-machine collaborative operation of drones can effectively improve the efficiency of pesticide use and ensure that each fruit tree is evenly sprayed with pesticides, thereby achieving better pest control effects.

[0003] Although drones have shown great potential for application in fruit tree spraying, the current spraying management method still has many shortcomings. Many orchards rely on fixed spraying paths during the spraying process and lack the ability to dynamically adjust to the growth heights and distribution densities of different fruit trees. This "one-size-fits-all" spraying strategy leads to uneven use of pesticides. Some fruit trees may not be adequately sprayed, while some areas may waste pesticides due to overlapping spraying. In addition, due to the lack of real-time data feedback and monitoring, the spraying effect is difficult to evaluate and it is difficult to make timely adjustments based on actual conditions. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for spraying fruit trees using two unmanned aerial vehicles in a coordinated intelligent manner, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for spraying fruit trees with two drones in a coordinated intelligent manner, comprising the following steps:

[0006] S1. Deploy laser radar and high-precision GPS on drones, use drone aerial photography to collect 3D point cloud data of orchards, and construct a 3D point cloud data set of orchards;

[0007] S2, preprocessing the data in the constructed orchard three-dimensional point cloud data set, cleaning the data and performing position correction;

[0008] S3. Based on the pre-processed 3D point cloud data of the orchard, a 3D grid model of the orchard is constructed, the positions of the fruit trees are marked, a set of task points for spraying pesticides in the orchard is constructed, and the orchard area is divided based on the height of the fruit trees to allocate the drone spraying tasks.

[0009] S4. For each drone, plan the drone spraying path based on the orchard 3D grid model, and assign the drone spraying path to two drones to perform the spraying operation. Use the sensors on the drone and the environmental monitoring device of the orchard to monitor the pesticide deposition and spraying effect in real time and collect data.

[0010] S5. Pre-set the pesticide spraying assessment threshold YZ, and obtain the comprehensive score YL of the orchard based on the real-time collected pesticide spraying volume Q, spraying uniformity JY and spraying efficiency E. By comparing and analyzing the orchard's comprehensive score YL with the preset pesticide spraying assessment threshold YZ, the current orchard pesticide spraying situation is evaluated and supplementary operations are performed.

[0011] Preferably, the specific steps of S1 include:

[0012] S11. Deploy lidar and high-precision GPS on drones, and set up ground control points (GCPS) with known precise coordinates in the orchard. Use drones equipped with lidar and high-precision GPS to take multiple aerial photos of the orchard to obtain three-dimensional point cloud data of the orchard, including orchard terrain data, fruit tree height data, and obstacle location data in the orchard. Merge the three-dimensional point cloud data of the orchard obtained from multiple aerial photos to construct a three-dimensional point cloud data set of the orchard.

[0013] Preferably, the specific steps of S2 include:

[0014] S21. Clean the orchard 3D point cloud data in the orchard 3D point cloud data set constructed in step S11, delete duplicate data points, identify and remove noise points and outliers, and convert all orchard 3D point cloud data into the same coordinate system. Combine the UAV GPS data and the GCPS coordinates of the ground control points in the orchard to correct the orchard 3D point cloud data and obtain a preprocessed orchard 3D point cloud data set.

[0015] Preferably, the specific steps of S3 include:

[0016] S31, using the modeling software MeshLab to convert the orchard 3D point cloud data set preprocessed in step S21 into an orchard 3D mesh model, and mark the position of each fruit tree in the orchard 3D mesh model, and record the position of each fruit tree as a pesticide spraying task point T i , get its coordinates (X i , Y i , Z i ), construct the orchard spraying task point set {T1, T2, ..., T n};

[0017] S32, randomly divide the orchard multiple times to obtain multiple different K values, where the K value refers to the number of divided areas of the orchard, obtain the average height of all fruit trees in each orchard area, and calculate the difference between the height of each fruit tree in the orchard area and the average height of the fruit trees, obtain the sum of squares of the errors in the orchard area, and thereby obtain the total sum of squares of the errors of the entire orchard area, plot the total sum of squares of the errors corresponding to different K values ​​into an SSE curve graph, obtain the SSE change rate between adjacent K values ​​based on the SSE curve graph, and use the K value with the smallest SSE change rate as the number of orchard area divisions;

[0018] S33. Based on the number K of orchard area divisions determined in S32, randomly select K heights from all the fruit tree heights in the orchard as cluster centers, calculate the Euclidean distance between each fruit tree and each cluster center, and assign each fruit tree to the cluster center with the smallest distance thereto, to form K orchard areas, whose Euclidean distance is obtained as follows:

[0019] d(h i , C j )=|h i -C j |;

[0020] In the formula, h i is the height of the i-th fruit tree, C k is the height of the cluster center of the jth orchard area;

[0021] S34. For each fruit tree in the orchard area, recalculate the cluster center of the orchard area, that is, the average height of all fruit trees in the orchard area.

[0022] S35, repeat the process of S34 until the cluster center of each orchard area no longer changes, so as to obtain the fruit tree position corresponding to each orchard area. Each orchard area corresponds to a cluster center, and a cluster center set {C1, C2, ..., C K}.

[0023] S36. According to the orchard clustering result of S35, K orchard areas and their corresponding mean heights of fruit trees are obtained, and the orchard areas are sorted according to the mean height of the fruit trees in each orchard area to obtain a list of orchard areas arranged from low to high in height. The orchard areas are allocated to each UAV in order from low to high, and each UAV obtains its own target area.

[0024] Preferably, the specific step of S4 includes:

[0025] S41, set the starting node of the drone spraying path as the initial position of the drone, spread over adjacent grids, search for the pesticide spraying task point, and calculate the total cost f(n) from the initial position of the drone to each surrounding pesticide spraying task point. The acquisition method is:

[0026] f(n)=g(n)+h(n);

[0027] Where g(n) is the actual cost from the initial position of the UAV to the task point of spraying the pesticide, and h(n) is the heuristic estimated cost from the initial position of the UAV to the task point of spraying the pesticide;

[0028] S42. According to step S41, the total cost f(n) from the initial position of the drone to each surrounding pesticide spraying task point is obtained, and a list of total costs f(n) is generated. The pesticide spraying task point with the smallest f(n) value is selected from the list as the next pesticide spraying task point to be reached by the drone. If the next pesticide spraying task point to be reached by the drone is the target area for which the drone is responsible, a path is constructed and the search is terminated. If the next pesticide spraying task point to be reached by the drone is not the target area for which the drone is responsible, the adjacent grids of the next pesticide spraying task point to be reached by the drone are traversed, and the total cost f(n) of the adjacent pesticide spraying task points from the initial position of the drone to the next pesticide spraying task point to be reached by the drone is calculated. The pesticide spraying task point with the smallest total cost f(n) is selected to determine whether it belongs to the target area for which the drone is responsible. The search is stopped until the next pesticide spraying task point to be reached by the drone is within the target area for which the drone is responsible.

[0029] S43: Set the selected pesticide spraying task point as the current drone task point, update the current position of the drone, repeat steps S41 and S42 until all target areas that the drone is responsible for are completed, and output the drone spraying path.

[0030] S44. Based on the UAV spraying path obtained in S43, the UAV is controlled to perform spraying operations. Before the UAV reaches the target area, the operating height of the UAV is controlled according to the average height of the fruit trees in the area obtained in S35 to perform spraying operations. The UAV spraying operation process is monitored by the sensors carried on the UAV and the orchard environment monitoring device, and the pesticide spraying amount Q, spraying uniformity JY, spraying efficiency E, total area, pesticide concentration and operation time of each pesticide spraying task point are recorded.

[0031] Preferably, the specific step of S5 includes:

[0032] S51, according to the spraying amount Q, spraying uniformity JY and spraying efficiency E of each spraying task point recorded in step S44, for the spraying task point T i , get the task point T of spraying the agent i Local spray effect score The local spray effect score Obtained through:

[0033]

[0034] In the formula, It is at the task point T of spraying the medicine i The coverage rate of the sprayed pesticide is It is the task point T for spraying medicine i The uniformity of spraying It is at the task point T of spraying the medicine i is the spraying efficiency of the UAV, α, β and γ are weight coefficients;

[0035] The above-mentioned agent spraying task point T i The coverage of the sprayed pesticide is CA Ti The method of obtaining is:

[0036]

[0037] In the formula, It is the task point T for spraying medicine i The amount of pesticide sprayed at It is the task point T for spraying medicine i The total area of ​​the above-mentioned pesticide spraying task point T i Spraying uniformity The method of obtaining is:

[0038]

[0039] In the formula, It is the task point T for spraying medicine i The standard deviation of the concentration of the drug at the random sampling point, It is the task point T for spraying medicine i The average value of the drug concentration at

[0040] The above-mentioned agent spraying task point T i The efficiency of drone spraying The method of obtaining is:

[0041]

[0042] In the formula, It is the task point T for spraying medicine i The time of drone spraying operation;

[0043] S52, based on the calculation of the task point T of the spraying agent in S51 i Local score of The local scores of n pesticide spraying task points in the orchard are obtained and weighted fusion is performed to obtain the comprehensive score YL of the orchard. The comprehensive score YL of the orchard is obtained as follows:

[0044]

[0045] In the formula, T i = T1, T2, T3, ..., T n , where n represents the number of medicine spraying task points in the orchard;

[0046] S53. Compare and analyze the comprehensive score YL of the orchard with the preset medicine spraying evaluation threshold YZ to evaluate the current medicine spraying situation in the orchard. The specific evaluation content is as follows:

[0047] If the comprehensive score YL of the orchard is greater than or equal to the preset medicine spraying evaluation threshold YZ, that is, when YL≥YZ, it indicates that the medicine spraying effect in the orchard is in a qualified state. Record the successful area of the orchard medicine spraying operation and store the medicine spraying strategy for this area.

[0048] If the comprehensive score YL of the orchard is less than the preset medicine spraying evaluation threshold YZ, that is, when YL < YZ, it indicates that the medicine spraying effect in the orchard is not in a qualified state. Trigger the alarm mechanism, generate an alarm message to notify the unmanned aerial vehicle operation dispatcher until a response is obtained.

[0049] S54. Generate an analysis report on the medicine spraying effect in the orchard based on the alarm message;

[0050] For the result that the medicine spraying effect in the orchard is not in a qualified state, collect the orchard medicine spraying data and control the unmanned aerial vehicle to perform supplementary operations.

[0051] The present invention provides a method for intelligent spraying of fruit trees by two cooperative unmanned aerial vehicles, which has the following beneficial effects:

[0052] (1) By carrying a lidar and a high-precision GPS on the unmanned aerial vehicle and combining with ground control points, the three-dimensional point cloud data of the orchard can be accurately obtained. These three-dimensional point cloud data not only cover detailed information such as the height of fruit trees, terrain, and obstacles, but also have undergone data cleaning and calibration processing to ensure the accuracy of the data. Based on these preprocessed data, the system can automatically construct a three-dimensional grid model and mark the positions of fruit trees. Through the fruit tree height clustering algorithm, the orchard is divided into multiple regions with similar heights. In this way, the fruit trees in different regions can be sprayed针对性, effectively reducing the waste of medicine caused by the height difference of fruit trees, while improving the spraying coverage rate and uniformity. The automated division of regions and allocation of spraying tasks improve the accuracy and overall efficiency of the unmanned aerial vehicle operation.

[0053] (2) Based on the three-dimensional grid model of the orchard, the spraying path of the drone is planned. The location data of the fruit trees in the three-dimensional grid model of the orchard and the heuristic cost estimation of the pesticide spraying task points are used to automatically select the operation path. When performing the task, the drone will operate point by point according to the planned path to reduce repeated flights and resource waste. In addition, the sensors on the drone and the environmental monitoring devices in the orchard are used to monitor the pesticide sedimentation and spraying effect in real time to ensure accurate control of the pesticide during the operation. The real-time monitoring system will dynamically adjust the operation status to ensure the coverage and uniformity of the pesticide spraying. At the same time, the feedback data is used to determine whether the spraying operation is qualified to ensure efficient and accurate orchard management.

[0054] (3) Through real-time evaluation of three key indicators, namely, the spraying volume Q, spraying uniformity JY and spraying efficiency E, the orchard comprehensive score YL is constructed to evaluate the overall spraying effect of the orchard. The orchard comprehensive score YL is compared with the preset spraying evaluation threshold YZ to accurately judge the quality of the current operation. When the orchard comprehensive score is lower than the spraying evaluation threshold YZ, the system will automatically generate an alarm message and prompt the operator to take additional operations to ensure that the spraying effect meets the standard. Through this intelligent evaluation and feedback mechanism, the spraying effect of the orchard is reliably guaranteed, while avoiding waste of resources and repeated operations, and improving the reliability and accuracy of drone spraying operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 The present invention is a flow chart of a method for the coordinated intelligent spraying of fruit trees using two unmanned aerial vehicles. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Example 1

[0058] See also Figure 1 The present invention provides a method for spraying fruit trees with two unmanned aerial vehicles in a coordinated intelligent manner, comprising the following steps:

[0059] S1. Deploy laser radar and high-precision GPS on drones, use drone aerial photography to collect 3D point cloud data of orchards, and construct a 3D point cloud data set of orchards;

[0060] S2, preprocessing the data in the constructed orchard three-dimensional point cloud data set, cleaning the data and performing position correction;

[0061] S3. Based on the pre-processed 3D point cloud data of the orchard, a 3D grid model of the orchard is constructed, the positions of the fruit trees are marked, a set of task points for spraying pesticides in the orchard is constructed, and the orchard area is divided based on the height of the fruit trees to allocate the drone spraying tasks.

[0062] S4. For each drone, plan the drone spraying path based on the orchard 3D grid model, and assign the drone spraying path to two drones to perform the spraying operation. Use the sensors on the drone and the environmental monitoring device of the orchard to monitor the pesticide deposition and spraying effect in real time and collect data.

[0063] S5. Pre-set the pesticide spraying assessment threshold YZ, and obtain the comprehensive score YL of the orchard based on the real-time collected pesticide spraying volume Q, spraying uniformity JY and spraying efficiency E. By comparing and analyzing the orchard's comprehensive score YL with the preset pesticide spraying assessment threshold YZ, the current orchard pesticide spraying situation is evaluated and supplementary operations are performed.

[0064] In this embodiment, by deploying laser radar and high-precision GPS, accurate collection and processing of orchard three-dimensional point cloud data can be achieved, and a three-dimensional grid model of the orchard can be dynamically constructed, and fruit trees can be highly clustered and regionalized. Compared with the traditional fixed spraying path, the present invention optimizes the spraying path planning of the drone according to the height difference and distribution density of the fruit trees, avoiding the waste of medicine or insufficient spraying caused by inconsistent tree heights. The spraying path can be accurately calculated between multiple task points, reducing the operation time of the drone and improving the spraying efficiency. In particular, the feedback mechanism for real-time monitoring of the settlement of the medicine and the spraying effect can dynamically evaluate the operation quality according to parameters such as the coverage rate of the medicine, the uniformity of the spraying and the efficiency, and make timely adjustments when necessary to ensure the uniformity and comprehensiveness of the spraying effect, greatly improving the utilization rate of the medicine. In addition, through the collaborative work of the two drones, tasks can be intelligently allocated in a large area of ​​orchards to ensure that fruit trees in different areas can obtain sufficient and uniform coverage of the medicine, solving the problems of low efficiency and uncontrollable effects in traditional spraying operations, and realizing the intelligent, precise and efficient management of orchard spraying.

[0065] Example 2

[0066] Please refer to Figure 1 ,Specific: S1 specific content includes,

[0067] S1 specific steps include:

[0068] S11. Deploy lidar and high-precision GPS on drones, and set up ground control points (GCPS) with known precise coordinates in the orchard. Use drones equipped with lidar and high-precision GPS to take multiple aerial photos of the orchard to obtain three-dimensional point cloud data of the orchard, including orchard terrain data, fruit tree height data, and obstacle location data in the orchard. Merge the three-dimensional point cloud data of the orchard obtained from multiple aerial photos to construct a three-dimensional point cloud data set of the orchard.

[0069] In this embodiment, by deploying lidar and high-precision GPS on the drone and setting the ground control point GCPS, the three-dimensional point cloud data of the orchard can be obtained with high precision. The use of lidar can realize accurate scanning of the orchard terrain, fruit tree height and obstacles. The multiple aerial photography and data merging of the drone can ensure comprehensive coverage of the entire orchard, eliminating the data omission problem caused by the limited viewing angle of traditional single aerial photography. Based on the precise coordinate calibration of GCPS, the accuracy and consistency of the data are significantly improved, ensuring the accurate positioning of the orchard terrain, fruit tree height and obstacle position. Compared with the traditional manual measurement method, the present invention greatly improves the efficiency and accuracy of data collection, and can provide a reliable data basis for subsequent orchard zoning, path planning and spraying task allocation, reducing uneven spraying or waste caused by inaccurate data, and improving the overall effect of the spraying operation.

[0070] Example 3

[0071] Please refer to Figure 1 ,Specifically: S2 includes:

[0072] S21. Clean the orchard 3D point cloud data in the orchard 3D point cloud data set constructed in step S11, delete duplicate data points, identify and remove noise points and outliers, and convert all orchard 3D point cloud data into the same coordinate system. Combine the UAV GPS data and the GCPS coordinates of the ground control points in the orchard to correct the orchard 3D point cloud data and obtain a preprocessed orchard 3D point cloud data set.

[0073] In this embodiment, by performing data cleaning on the orchard three-dimensional point cloud data set, duplicate data points can be effectively deleted, noise points and outliers can be removed, and the accuracy and reliability of the data can be ensured. This data preprocessing process not only greatly improves the accuracy of the data, but also reduces error accumulation, avoiding unnecessary interference to subsequent analysis and operations due to abnormal data. By converting all three-dimensional point cloud data into a unified coordinate system and combining the GPS data of the drone and the precise coordinates of the ground control point GCPS for correction, the spatial consistency and high-precision positioning of all point cloud data in the orchard can be ensured, avoiding deviations caused by inconsistent coordinates in traditional data collection. The preprocessed point cloud data provides a solid data foundation for the subsequent construction of the orchard three-dimensional grid model, fruit tree zoning, and drone spraying path planning, ensuring the accuracy and efficiency of drone spraying, reducing the waste of pesticides, and achieving accurate and uniform spraying of fruit trees, thereby improving the management level and resource utilization of the orchard.

[0074] Example 4

[0075] Please refer to Figure 1 ,Specifically: S3 includes:

[0076] S31, using the modeling software MeshLab to convert the orchard 3D point cloud data set preprocessed in step S21 into an orchard 3D mesh model, and mark the position of each fruit tree in the orchard 3D mesh model, and record the position of each fruit tree as a pesticide spraying task point T i , get its coordinates (X i , Y i , Z i ), construct the orchard spraying task point set {T1, T2, ..., T n};

[0077] S32, randomly divide the orchard multiple times to obtain multiple different K values, where the K value refers to the number of divided areas of the orchard, obtain the average height of all fruit trees in each orchard area, and calculate the difference between the height of each fruit tree in the orchard area and the average height of the fruit trees, obtain the sum of squares of the errors in the orchard area, and thereby obtain the total sum of squares of the errors of the entire orchard area, plot the total sum of squares of the errors corresponding to different K values ​​into an SSE curve graph, obtain the SSE change rate between adjacent K values ​​based on the SSE curve graph, and use the K value with the smallest SSE change rate as the number of orchard area divisions;

[0078] S33. Based on the number K of orchard area divisions determined in S32, randomly select K heights from all the fruit tree heights in the orchard as cluster centers, calculate the Euclidean distance between each fruit tree and each cluster center, and assign each fruit tree to the cluster center with the smallest distance thereto, to form K orchard areas, whose Euclidean distance is obtained as follows:

[0079] d(h i , C j )=|h i -C j |;

[0080] In the formula, h i is the height of the i-th fruit tree, C k is the height of the cluster center of the jth orchard area;

[0081] S34. For each fruit tree in the orchard area, recalculate the cluster center of the orchard area, that is, the average height of all fruit trees in the orchard area.

[0082] S35, repeat the process of S34 until the cluster center of each orchard area no longer changes, so as to obtain the fruit tree position corresponding to each orchard area. Each orchard area corresponds to a cluster center, and a cluster center set {C1, C2, ..., C K}.

[0083] S36. According to the orchard clustering result of S35, K orchard areas and their corresponding mean heights of fruit trees are obtained, and the orchard areas are sorted according to the mean height of the fruit trees in each orchard area to obtain a list of orchard areas arranged from low to high in height. The orchard areas are allocated to each UAV in order from low to high, and each UAV obtains its own target area.

[0084] In this embodiment, the pre-processed orchard 3D point cloud data set is converted into an orchard 3D grid model, the position of each fruit tree is accurately marked, and a set of orchard pesticide spraying task points is constructed, which not only provides the precise coordinates (X i , Y i , Z i), and also provide accurate navigation reference for UAV in actual operation, ensuring the coverage of pesticide spraying task points without omission, greatly improving the accuracy of spraying operation, and ensuring the path planning and operation efficiency of UAV through accurate division of orchard area and accurate positioning of each fruit tree. The orchard is divided into regions based on the mean height of fruit trees, and the selection of K value is optimized through the SSE error square sum curve to ensure that the height difference of each orchard area is minimized. This adaptive division method can dynamically adjust the area size according to the distribution of fruit tree height, avoid the "one-size-fits-all" regional division strategy, and effectively solve the traditional spraying method due to The problem of pesticide waste or insufficient coverage caused by uneven heights of fruit trees. In addition, by calculating the distance between each fruit tree and its corresponding cluster center, the heights of fruit trees in the orchard are ensured to be similar, and the position of the cluster center is continuously optimized to finally obtain a stable orchard area division result. This process can make the drone's spraying task more accurately match the growth of fruit trees. The area assigned to each drone is also carefully calculated and sorted to ensure that the drone's operating height matches the height of the fruit trees, thereby improving the coverage and uniformity of the spray, reducing the waste of pesticides caused by height differences, and achieving efficient utilization of resources and precise management of orchard spraying.

[0085] Example 5

[0086] Please refer to Figure 1 ,Specific: S4 specific content includes,

[0087] S41, set the starting node of the drone spraying path as the initial position of the drone, spread over adjacent grids, search for the pesticide spraying task point, and calculate the total cost f(n) from the initial position of the drone to each surrounding pesticide spraying task point. The acquisition method is:

[0088] f(n)=g(n)+h(n);

[0089] Where g(n) is the actual cost from the initial position of the UAV to the task point of spraying the pesticide, and h(n) is the heuristic estimated cost from the initial position of the UAV to the task point of spraying the pesticide;

[0090] The actual cost from the initial position of the UAV to the task point of spraying the pesticide is that there are obstacles in the orchard. When planning the flight path of the UAV, it is necessary to avoid the obstacles. The heuristic estimated cost from the initial position of the UAV to the task point of spraying the pesticide refers to the straight-line distance from the UAV to the task point of spraying the pesticide.

[0091] S42. According to step S41, the total cost f(n) from the initial position of the drone to each surrounding pesticide spraying task point is obtained, and a list of total costs f(n) is generated. The pesticide spraying task point with the smallest f(n) value is selected from the list as the next pesticide spraying task point to be reached by the drone. If the next pesticide spraying task point to be reached by the drone is the target area for which the drone is responsible, a path is constructed and the search is terminated. If the next pesticide spraying task point to be reached by the drone is not the target area for which the drone is responsible, the adjacent grids of the next pesticide spraying task point to be reached by the drone are traversed, and the total cost f(n) of the adjacent pesticide spraying task points from the initial position of the drone to the next pesticide spraying task point to be reached by the drone is calculated. The pesticide spraying task point with the smallest total cost f(n) is selected to determine whether it belongs to the target area for which the drone is responsible, and the search is stopped until the next pesticide spraying task point to be reached by the drone is within the target area for which the drone is responsible.

[0092] S43: Set the selected pesticide spraying task point as the current drone task point, update the current position of the drone, repeat steps S41 and S42 until all target areas that the drone is responsible for are completed, and output the drone spraying path.

[0093] S44, based on the drone spraying path obtained in S43, control the drone to perform spraying operations. Before the drone reaches the target area, control the drone's operating height according to the average height of the fruit trees in the area obtained in S35 to perform spraying operations, and monitor the drone's spraying operation process through sensors and orchard environment monitoring devices carried on the drone, and record the pesticide spraying amount Q, spraying uniformity JY, spraying efficiency E, total area, pesticide concentration and operation time of each pesticide spraying task point;

[0094] The above-mentioned sensor carried on the drone refers to a flow sensor, which is used to monitor the flow of liquid medicine from the drone nozzle. The orchard environment monitoring device refers to an electrochemical sensor. The total area of ​​the pesticide spraying task point is obtained through the orchard three-dimensional grid model, the pesticide concentration is obtained through the electrochemical sensor, and the operation time is obtained through the drone operation data in the drone control system.

[0095] In this implementation, the starting node of the drone's spraying path is set as the drone's initial position, and the pesticide spraying task point is searched by spreading over adjacent grids. The total cost from the drone's initial position to each pesticide spraying task point is calculated. The actual cost and the heuristic estimated cost are combined to ensure that the drone can spray efficiently based on the path. This cost calculation method not only optimizes the drone's flight route, but also effectively reduces the time and resource waste caused by improper path selection during the spraying process, laying the foundation for subsequent spraying operations. A total cost list is generated, and the pesticide spraying task point with the smallest value is selected as the drone's next target point to ensure that the drone always moves towards the target area. If the next pesticide spraying task point that the drone arrives at is the orchard area it is responsible for, a spraying path will be constructed and the search will end. Such a dynamic selection mechanism enables the drone to Adaptive adjustment of spraying strategies in complex environments avoids low operating efficiency and waste of pesticides caused by fixed paths, ensuring that spraying tasks are carried out efficiently and accurately. As the drone continuously updates its current position and repeats the selection of task points, it finally outputs the spraying path. The continuous optimization of this process can adjust the drone's spraying strategy in real time, improving the accuracy and efficiency of spraying operations. By controlling the height of the drone during spraying operations and dynamically adjusting it based on the average height of the fruit trees, it ensures that the pesticide can be evenly sprayed on each fruit tree. In addition, the onboard sensors and monitoring devices are used to record the pesticide spraying volume Q, spraying uniformity JY and spraying efficiency E of each pesticide spraying task point, providing reliable data support for subsequent spraying effect evaluation, thereby achieving the goal of precision agriculture and further improving the scientificity and effectiveness of orchard spraying operations.

[0096] Example 6

[0097] Please refer to Figure 1 ,Specific: S5 specific content includes,

[0098] S51, according to the spraying amount Q, spraying uniformity JY and spraying efficiency E of each spraying task point recorded in step S44, for the spraying task point T i , get the task point T of spraying the agent i Local spray effect score The local spray effect score Obtained through:

[0099]

[0100] In the formula, It is at the task point T of spraying the medicine i The coverage rate of the sprayed pesticide is It is the task point T for spraying medicine i The uniformity of spraying It is at the task point T of spraying the medicinei The spraying efficiency of the UAV is represented by α, β and γ, which are weight coefficients α+β+γ=1;

[0101] The above-mentioned agent spraying task point T i Coverage of sprayed pesticide The method of obtaining is:

[0102]

[0103] In the formula, It is the task point T for spraying medicine i The amount of pesticide sprayed at It is the task point T for spraying medicine i The total area of ​​the above-mentioned pesticide spraying task point T i Spraying uniformity JY Ti The method of obtaining is:

[0104]

[0105] In the formula, It is the task point T for spraying medicine i The standard deviation of the concentration of the drug at the random sampling point, It is the task point T for spraying medicine i The average value of the drug concentration at

[0106] The standard deviation and average value of the above-mentioned concentration of the drug refer to the concentration at the task point T of the drug spraying. i Randomly select sampling points at the locations to sample the concentration of the agent and obtain the task point T for spraying the agent. i The standard deviation and mean of the drug concentration at

[0107] The above-mentioned agent spraying task point T i The efficiency of drone spraying The method of obtaining is:

[0108]

[0109] In the formula, It is the task point T for spraying medicine i The time for drone spraying operations.

[0110] S52, based on the calculation of the task point T of the spraying agent in S51 i Local score of The local scores of n pesticide spraying task points in the orchard are obtained and weighted fusion is performed to obtain the comprehensive score YL of the orchard. The comprehensive score YL of the orchard is obtained as follows:

[0111]

[0112] Wherein, T i = T1, T2, T3, ..., T n , where n represents the number of medicament spraying task points in the orchard.

[0113] S53. Compare and analyze the comprehensive score YL of the orchard obtained in step S52 with the preset medicament spraying evaluation threshold YZ to evaluate the current medicament spraying situation in the orchard. The specific evaluation content is as follows:

[0114] If the comprehensive score YL of the orchard is greater than or equal to the preset medicament spraying evaluation threshold YZ, that is, when YL≥YZ, it indicates that the medicament spraying effect in the orchard is in a qualified state. Record the successful area of the orchard spraying operation and store the spraying strategy for this area.

[0115] If the comprehensive score YL of the orchard is less than the preset medicament spraying evaluation threshold YZ, that is, when YL<YZ, it indicates that the medicament spraying effect in the orchard is not in a qualified state. Trigger the alarm mechanism, generate an alarm message to notify the drone operation dispatcher until a response is obtained.

[0116] S54. Generate an analysis report on the medicament spraying effect in the orchard based on the alarm message.

[0117] For the result that the medicament spraying effect in the orchard is not in a qualified state, collect the medicament spraying data in the orchard and control the drone to perform supplementary operations.

[0118] In this embodiment, by recording the medicament spraying amount Q, spraying uniformity JY, and spraying efficiency E of each medicament spraying task point, for each medicament spraying task point T iGenerate a local spraying effect score. By introducing a weight coefficient, the evaluation can be more flexible and targeted. The quantitative analysis of the pesticide coverage, spraying uniformity and spraying efficiency helps to accurately evaluate the spraying effect of each task point. This detailed local evaluation helps to identify possible problems in the spraying process and provides a data basis for subsequent optimization to ensure that the spraying operation is more efficient. The local scores of all pesticide spraying task points are weighted and integrated to obtain the orchard's comprehensive score YL. This comprehensively considers the individual performance of each pesticide spraying task point and integrates the spraying effect of the entire orchard, thereby providing a more comprehensive effect evaluation, which can help decision makers better understand the overall spraying effect, identify potential weaknesses and room for improvement, and promote the continuous optimization of the spraying strategy. In addition, the orchard's comprehensive score YL is compared and analyzed with the preset pesticide spraying evaluation threshold YZ to quickly judge the eligibility of the current spraying effect. If the orchard's comprehensive score YL is higher than or equal to the pesticide spraying threshold YZ, the orchard's comprehensive score YL can be used to evaluate the spraying effect. The spraying evaluation threshold YZ indicates that the spraying effect is qualified. At this time, the successful area is recorded and the relevant spraying strategies are stored to ensure the sustainability and effectiveness of future spraying operations. If the comprehensive score YL of the orchard is lower than the pesticide spraying evaluation threshold YZ, the alarm mechanism is automatically triggered and the operation dispatching personnel are notified, so that timely measures can be taken to avoid further waste of resources and operation failure. This dynamic feedback mechanism improves the flexibility and response ability of spraying operations, and provides important support for orchard management. Finally, an orchard pesticide spraying effect analysis report is generated, which summarizes the data and effects of spraying operations, can provide a clear decision-making basis for orchard management, and provide a reference for the improvement of future spraying operations. In addition, when the spraying effect is not up to standard, data can be quickly collected to assist drone dispatchers in controlling drones for supplementary operations, ensuring the healthy growth of crops and improving the efficiency of drone spraying operations. It also realizes the comprehensive optimization of orchard management and provides strong support for the development of smart agriculture.

[0119] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for spraying fruit trees with two drones in a coordinated intelligent manner, characterized in that: The following steps are included: S1. Deploy laser radar and high-precision GPS on drones, use drone aerial photography to collect 3D point cloud data of orchards, and construct a 3D point cloud data set of orchards; S2, preprocessing the data in the constructed orchard three-dimensional point cloud data set, cleaning the data and performing position correction; S3. Based on the pre-processed 3D point cloud data of the orchard, a 3D grid model of the orchard is constructed, the positions of the fruit trees are marked, a set of task points for spraying pesticides in the orchard is constructed, and the orchard area is divided based on the height of the fruit trees to allocate the drone spraying tasks. S4. For each drone, plan the drone spraying path based on the orchard 3D grid model, and assign the drone spraying path to two drones to perform the spraying operation. Use the sensors on the drone and the environmental monitoring device of the orchard to monitor the pesticide deposition and spraying effect in real time and collect data. S5. Pre-set the pesticide spraying assessment threshold YZ, and obtain the comprehensive score YL of the orchard based on the real-time collected pesticide spraying volume Q, spraying uniformity JY and spraying efficiency E. By comparing and analyzing the orchard's comprehensive score YL with the preset pesticide spraying assessment threshold YZ, the current orchard pesticide spraying situation is evaluated and supplementary operations are performed.

2. The method for spraying pesticides by using two drones in a coordinated intelligent manner for fruit trees according to claim 1, characterized in that: S1 specific steps include: S11. Deploy lidar and high-precision GPS on drones, and set up ground control points (GCPS) with known precise coordinates in the orchard. Use drones equipped with lidar and high-precision GPS to take multiple aerial photos of the orchard to obtain three-dimensional point cloud data of the orchard, including orchard terrain data, fruit tree height data, and obstacle location data in the orchard. Merge the three-dimensional point cloud data of the orchard obtained from multiple aerial photos to construct a three-dimensional point cloud data set of the orchard.

3. The method for spraying pesticides by using two drones in a coordinated intelligent manner for fruit trees according to claim 2, characterized in that: The specific steps of S2 include: S21. Clean the orchard 3D point cloud data in the orchard 3D point cloud data set constructed in step S11, delete duplicate data points, identify and remove noise points and outliers, and convert all orchard 3D point cloud data into the same coordinate system. Combine the UAV GPS data and the GCPS coordinates of the ground control points in the orchard to correct the orchard 3D point cloud data and obtain a preprocessed orchard 3D point cloud data set.

4. The method for spraying fruit trees with two drones in a coordinated intelligent manner according to claim 3 is characterized in that: The specific steps of S3 include: S31, using the modeling software MeshLab to convert the orchard 3D point cloud data set preprocessed in step S21 into an orchard 3D mesh model, and mark the position of each fruit tree in the orchard 3D mesh model, and record the position of each fruit tree as a pesticide spraying task point T i , get its coordinates (X i , Y i , Z i ), construct the orchard spraying task point set {T1, T2, ..., T n }; S32, randomly divide the orchard multiple times to obtain multiple different K values, where the K value refers to the number of divided areas of the orchard, obtain the average height of all fruit trees in each orchard area, and calculate the difference between the height of each fruit tree in the orchard area and the average height of the fruit trees, obtain the sum of squares of the errors in the orchard area, and thereby obtain the total sum of squares of the errors of the entire orchard area, plot the total sum of squares of the errors corresponding to different K values ​​into an SSE curve graph, obtain the SSE change rate between adjacent K values ​​based on the SSE curve graph, and use the K value with the smallest SSE change rate as the number of orchard area divisions; S33. Based on the number K of orchard area divisions determined in S32, randomly select K heights from all the fruit tree heights in the orchard as cluster centers, calculate the Euclidean distance between each fruit tree and each cluster center, and assign each fruit tree to the cluster center with the smallest distance thereto, to form K orchard areas, whose Euclidean distance is obtained as follows: d(h i ,C j )=|h i -C j |; In the formula, h i is the height of the i-th fruit tree, C k is the height of the cluster center of the jth orchard area; S34. For each fruit tree in the orchard area, recalculate the cluster center of the orchard area, that is, the average height of all fruit trees in the orchard area.

5. The method for spraying fruit trees with two drones in a coordinated intelligent manner according to claim 4, characterized in that: S35, repeat the process of S34 until the cluster center of each orchard area no longer changes, so as to obtain the fruit tree position corresponding to each orchard area. Each orchard area corresponds to a cluster center, and a cluster center set {C1, C2, ..., C K }.

6. The method for spraying pesticides with two drones in coordination with each other for fruit trees according to claim 5 is characterized in that: S36. According to the orchard clustering result of S35, K orchard areas and their corresponding mean heights of fruit trees are obtained, and the orchard areas are sorted according to the mean height of the fruit trees in each orchard area to obtain a list of orchard areas arranged from low to high in height. The orchard areas are allocated to each UAV in order from low to high, and each UAV obtains its own target area.

7. The method for spraying pesticides with two drones in coordination with each other for fruit trees according to claim 6 is characterized by: The specific steps of S4 include: S41, set the starting node of the drone spraying path as the initial position of the drone, spread over adjacent grids, search for the pesticide spraying task point, and calculate the total cost f(n) from the initial position of the drone to each surrounding pesticide spraying task point. The acquisition method is: f(n)=g(n)+h(n); Where g(n) is the actual cost from the initial position of the UAV to the task point of spraying the pesticide, and h(n) is the heuristic estimated cost from the initial position of the UAV to the task point of spraying the pesticide; S42. Obtain the total cost f(n) from the initial position of the UAV to each surrounding pesticide spraying task point according to step S41, generate a list of the total cost f(n), and select the pesticide spraying task point with the minimum f(n) value as the next pesticide spraying task point for the UAV to reach. If the next pesticide spraying task point for the UAV to reach is the target area responsible for the UAV, construct a path and end the search. If the next pesticide spraying task point for the UAV to reach is not the target area responsible for the UAV, traverse the adjacent grids of the next pesticide spraying task point for the UAV to reach, calculate the total cost f(n) from the initial position of the UAV to the adjacent pesticide spraying task points of the next pesticide spraying task point for the UAV to reach, select the pesticide spraying task point with the minimum total cost f(n), and judge whether it belongs to the target area responsible for the UAV until the next pesticide spraying task point for the UAV to reach is within the target area responsible for the UAV, and then stop the search; S43. Take the selected pesticide spraying task point as the current UAV task point, update the current position of the UAV, and repeat steps S41 and S42 until all the target areas responsible for the UAV are completed, and output the UAV spraying path.

8. A method for dual-UAV collaborative intelligent pesticide spraying for fruit trees according to claim 7, characterized in that: S44. Based on the UAV spraying path obtained in S43, control the UAV to perform pesticide spraying operations. Before the UAV reaches the target area, control the operating height of the UAV according to the average height of the fruit trees in the area obtained in S35 to perform pesticide spraying operations, and monitor the UAV pesticide spraying operation process through the sensors and orchard environment monitoring devices carried on the UAV, and record the pesticide spraying amount Q, spraying uniformity JY, spraying efficiency E, total area, pesticide concentration and operation time of each pesticide spraying task point.

9. A method for dual-UAV collaborative intelligent pesticide spraying for fruit trees according to claim 8, characterized in that: The specific steps of S5 include: S51, according to the spraying amount Q, spraying uniformity JY and spraying efficiency E of each spraying task point recorded in step S44, for the spraying task point T i , get the task point T of spraying the agent i Local spray effect score The local spray effect score Obtained through: In the formula, It is at the task point T of spraying the medicine i The coverage rate of the sprayed pesticide is It is the task point T for spraying medicine i The uniformity of spraying It is at the task point T of spraying the medicine i is the spraying efficiency of the UAV, α, β and γ are weight coefficients; The above-mentioned agent spraying task point T i Coverage of sprayed pesticide The method of obtaining is: In the formula, It is the task point T for spraying medicine i The amount of pesticide sprayed at It is the task point T for spraying medicine i The total area of ​​the place; The above-mentioned agent spraying task point T i Spraying uniformity The method of obtaining is: In the formula, It is the task point T for spraying medicine i The standard deviation of the concentration of the drug at the random sampling point, It is the task point T for spraying medicine i The average value of the drug concentration at The above-mentioned agent spraying task point T i The efficiency of spraying by drones The method of obtaining is: In the formula, It is the task point T for spraying medicine i The time of drone spraying operation; S52, based on the calculation of the task point T of the spraying agent in S51 i Local score of The local scores of n pesticide spraying task points in the orchard are obtained and weighted fusion is performed to obtain the comprehensive score YL of the orchard. The comprehensive score YL of the orchard is obtained as follows: Where, T i =T1, T2, T3, ..., T n , n represents the number of pesticide spraying task points in the orchard.

10. A method for dual-UAV collaborative intelligent pesticide spraying for fruit trees according to claim 1, characterized in that: S53. Compare and analyze the comprehensive score YL of the orchard with the preset pesticide spraying evaluation threshold YZ to evaluate the current pesticide spraying situation in the orchard. The specific evaluation content is as follows: If the comprehensive score YL of the orchard is greater than or equal to the preset pesticide spraying evaluation threshold YZ, that is, when YL≥YZ, it means that the pesticide spraying effect in the orchard is in a qualified state, record the successful area of the orchard pesticide spraying operation, and store the pesticide spraying strategy for this area; If the comprehensive score YL of the orchard is less than the preset pesticide spraying evaluation threshold YZ, that is, when YL<YZ, it means that the pesticide spraying effect in the orchard is not in a qualified state, trigger the alarm mechanism, generate an alarm message to notify the UAV operation dispatcher until a response is obtained; S54. Generate an analysis report on the pesticide spraying effect of the orchard based on the alarm message; For the result that the pesticide spraying effect in the orchard is not in a qualified state, collect the orchard pesticide spraying data and control the UAV to perform supplementary operations.

Images