Large-crown fruit tree spraying operation control optimization method, device, equipment and medium

Abstract

The invention relates to a large-crown fruit tree spraying operation control optimization method, device and equipment and a medium, and the method comprises the steps: inputting the canopy porosity and the boundary coordinates of a target region into a spraying operation control model, so as to determine an initial spraying site corresponding to the initial flight parameters of a plant protection unmanned aerial vehicle; the plant protection unmanned aerial vehicle is controlled to fly to the initial spraying site to hover, the hovering site of the plant protection unmanned aerial vehicle is gradually and finely adjusted in the transverse axis direction and the longitudinal axis direction, the characteristic parameters of the disturbance area are detected again after adjustment every time until the disturbance area corresponding to the characteristic parameters of the disturbance area coincides with the target area, and first spraying operation is conducted on the target area; detecting fogdrop deposition characteristic parameters corresponding to the target area, and if the fogdrop deposition characteristic parameters do not reach a preset threshold value, performing second spraying operation on the target area with insufficient deposition until the fogdrop deposition characteristic parameters reach the preset threshold value. The problem that the disturbance area is difficult to match with the target area due to the complex canopy structure of the large-crown fruit tree is solved.

Description

Technical Field

[0001] This application relates to the field of drone operations, and in particular to a method for controlling and optimizing spraying operations on large-canopy fruit trees, a corresponding device, electronic equipment, and a computer-readable storage medium. Background Technology

[0002] With the trend of precision development in modern agriculture, plant protection drones have been widely used in fruit tree control operations due to their high efficiency, flexibility and environmental friendliness.

[0003] Currently, for large-canopy fruit trees (such as lychee, longan, citrus, and mango), due to their tall stature and dense canopy, the spray width and rotor airflow disturbance of existing plant protection drone spraying systems are limited, making it difficult to achieve uniform coverage of the entire canopy surface and interior. This results in unsatisfactory control effects, with prominent problems such as insufficient droplet deposition and uneven target coverage.

[0004] In summary, existing plant protection drone spraying systems have limited spray width and rotor airflow disturbance, making it difficult to achieve uniform coverage of the entire canopy surface and interior. This results in unsatisfactory control effects, insufficient droplet deposition, and uneven target coverage. The applicant has made corresponding explorations to address these issues. Summary of the Invention

[0005] The purpose of this application is to solve the above-mentioned problems by providing a method for optimizing the control of spraying operations on large-canopy fruit trees, a corresponding device, electronic equipment, and a computer-readable storage medium.

[0006] To achieve the various objectives of this application, the following technical solution is adopted: A method for optimizing the control of spraying operations on large-canopy fruit trees, proposed to meet one of the purposes of this application, includes: Based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, the canopy structure information and target area of ​​the fruit tree to be sprayed are determined. The canopy structure information includes canopy diameter, canopy height and canopy porosity. In the two-dimensional coordinate system of the fruit tree canopy, the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters are input into the spraying operation control model that has been trained to a convergent state to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by a mapping function from flight parameters to disturbance area feature parameters and a mapping function from disturbance area feature parameters to droplet deposition feature parameters. The agricultural drone is controlled to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy layer is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy are determined. The hovering position of the agricultural drone is gradually fine-tuned along the horizontal and vertical axes. After each adjustment, the characteristic parameters of the disturbed area are re-detected until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to carry out the first spraying operation on the target area. The droplet deposition characteristic parameters corresponding to the target area are detected. If the droplet deposition characteristic parameters do not reach a preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is called to perform reverse calculation based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

[0007] Optionally, the flight parameters characterize the flight status parameters of the agricultural drone during spraying operations, and are used to control the spatial position and rotor power output of the agricultural drone. They include rotor speed, hovering height, horizontal axis coordinate, and horizontal axis coordinate. The boundary coordinates of the target region represent the coordinates of the diagonal vertices of the smallest bounding rectangle of the target region in the canopy two-dimensional coordinate system; The canopy porosity is a parameter that characterizes the density of branches and leaves inside the canopy of a fruit tree. The canopy porosity includes the average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone. The average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone correspond to the proportion of gaps between branches and leaves in different areas of the fruit tree canopy. The environmental parameters characterize the external environmental conditions related to the spraying operation, including ambient wind speed and ambient temperature. The spray system parameters characterize the working status parameters of the spray system of the agricultural drone, including spray pressure and atomized particle size; The disturbance region characteristic parameters characterize the spatial attribute parameters of the branch and leaf disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy. These parameters include the coordinates of the center point in the canopy two-dimensional coordinate system, the area of ​​the minimum circumscribed rectangle, and the angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system. The droplet deposition characteristic parameters characterize the adhesion effect of droplets in the target area after spraying, including the droplet deposition density and droplet coverage in the target area.

[0008] Optional steps for training the spraying operation control model include: Obtain a sample dataset, wherein the sample dataset includes multiple sets of sample data, the sample data includes input data and output data, the input data includes flight parameters of the agricultural drone, canopy porosity, environmental parameters and spray system parameters; the output data includes disturbance area characteristic parameters and droplet deposition characteristic parameters; The parameters of the spraying system are fixed, and the flight parameters, canopy porosity and environmental parameters of the agricultural drone are adjusted to collect the characteristic parameters of the disturbed area corresponding to different parameter combinations. A neural network regression algorithm is used to iteratively optimize the data with the goal of minimizing the first mean square error between the predicted and actual values ​​of the characteristic parameters of the disturbed area until the first mean square error reaches below a preset threshold, so as to determine the mapping function from the flight parameters to the characteristic parameters of the disturbed area. The flight parameters are fixed, and the characteristic parameters of the disturbed area, the canopy porosity, and the spray system parameters are adjusted. The sample data of the low penetration area is supplemented in particular to collect the droplet deposition characteristic parameters corresponding to different parameter combinations. A neural network regression algorithm is used to add the error weight of the correspondence between the characteristic parameters of the disturbed area and the characteristic parameters of the target area to the preset loss function. The algorithm is iteratively optimized with the goal of minimizing the second mean square error between the predicted and actual values ​​of the droplet deposition characteristic parameters until the second mean square error value reaches below the preset threshold, so as to determine the mapping function from the characteristic parameters of the disturbed area to the droplet deposition characteristic parameters. The mapping functions from flight parameters to disturbance area characteristic parameters and from disturbance area characteristic parameters to droplet deposition characteristic parameters are integrated to construct an end-to-end spraying operation control model. Select a portion of sample data as a test set to verify the model accuracy. If the prediction deviation of the perturbation area feature parameters and droplet deposition feature parameters exceeds the preset threshold, supplement the corresponding sample data and retrain the mapping function until the spraying operation control model meets the preset accuracy.

[0009] Optionally, after determining the canopy structure information and target area of ​​the fruit trees to be sprayed based on the canopy image of the fruit trees to be sprayed and its corresponding three-dimensional point cloud data, the process includes: A two-dimensional coordinate system for the fruit tree canopy is constructed with a preset location point as the origin. The structural information of the fruit tree canopy is mapped to the two-dimensional coordinate system for modeling to determine the canopy plane area. The coordinate boundary of the target area is marked on the two-dimensional coordinate system.

[0010] Optionally, the step of extracting the dynamic region of the fruit tree canopy image using the inter-frame difference method and extracting the two-dimensional projection point cloud of the three-dimensional point cloud data onto the fruit tree canopy layers, and determining the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy based on the spatial displacement changes of the dynamic region and the two-dimensional projection point cloud, includes: Control the plant protection drone that has flown to the initial spraying site and hovers to continuously collect images of the canopy of the fruit trees to be sprayed, as well as three-dimensional point cloud data that is spatiotemporally synchronized with the canopy images of the fruit trees. After preprocessing the continuously acquired fruit tree canopy images, the inter-frame difference method is used to calculate the pixel gray value difference between two adjacent frames of the fruit tree canopy images to generate a difference image. The pixel areas in the difference image where the gray value change rate exceeds the preset change rate threshold are identified as the dynamic areas of the fruit tree canopy affected by the rotor airflow. Based on the canopy height of the fruit trees to be sprayed, the three-dimensional point cloud data is processed into layers of the fruit tree canopy at preset height intervals. The two-dimensional projection point cloud of each layer of three-dimensional point cloud data in the two-dimensional coordinate system of the canopy is extracted to obtain the two-dimensional projection point cloud corresponding to each layer. A time series analysis is performed on the two-dimensional projected point cloud of each layer to calculate the spatial displacement and point cloud density change of the two-dimensional projected point cloud per unit time; the two-dimensional projected point cloud regions where the spatial displacement exceeds a preset displacement threshold or the point cloud density change exceeds a preset density change threshold are identified as candidate disturbance regions. The dynamic region is spatially matched with the candidate disturbance regions of each layer, and the candidate disturbance regions that spatially overlap with the dynamic region are selected as the effective disturbance regions formed by the interaction between the rotor airflow and the fruit tree canopy. The center point coordinates, the area of ​​the minimum circumscribed rectangle, and the angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system of the effective disturbance region are extracted to construct the characteristic parameters of the disturbance region.

[0011] Optionally, the hovering position of the agricultural drone is gradually fine-tuned along the horizontal and vertical axes, and the characteristic parameters of the disturbed area are re-detected after each adjustment until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to perform the first spraying operation on the target area. This step includes: Based on the canopy two-dimensional coordinate system, the spatial position of the disturbance region corresponding to the disturbance region feature parameters is compared with that of the target region to determine whether the center point of the disturbance region coincides with the center point of the target region; if they do not coincide, the first deviation value of the center point of the disturbance region relative to the center point of the target region in the horizontal axis direction and the second deviation value in the vertical axis direction of the canopy two-dimensional coordinate system are calculated. Based on the first deviation value and the second deviation value, the hovering position of the agricultural drone is gradually fine-tuned along the horizontal axis and the vertical axis by a preset step size, wherein the preset fine-tuning step size is less than or equal to 0.1m; After each adjustment of the hovering position, the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data are re-acquired. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy are determined. Repeat the above steps until the overlap between the perturbation region corresponding to the re-detected perturbation region feature parameters and the target region in the canopy two-dimensional coordinate system reaches a preset percentage, and the area of ​​the perturbation region is greater than or equal to a certain multiple of the target region area; maintain the current hovering position, start the spraying system of the plant protection drone, and perform the first spraying operation on the target region according to the initially set spraying pressure and atomized particle size parameters, wherein the preset percentage is greater than or equal to 95%, and the certain multiple is greater than the value 1.

[0012] Optionally, the step of calling the spraying operation control model to perform reverse calculation based on the boundary coordinates of the insufficient deposition target area to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site, and adjusting the spraying system parameters of the agricultural drone according to the canopy porosity corresponding to the insufficient deposition target area to perform a second spraying operation on the insufficient deposition target area until the droplet deposition characteristic parameters reach a preset threshold includes: If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the high penetration area, the operating height of the agricultural drone will be reduced by 5% to 10%. If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the medium penetration zone, the spray pressure of the spraying system will be increased by 10% to 15%, while reducing the operating height of the agricultural drone by 5% to 10%. If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the low penetration area, in addition to reducing the operating height of the agricultural drone by 5% to 10% and increasing the spray pressure of the spray system by 10% to 15%, the atomized particle size of the nozzle of the spray system will be reduced by 20% to 30%.

[0013] A large-canopy fruit tree spraying operation control and optimization device provided for another purpose of this application includes: The canopy image acquisition module is configured to determine the canopy structure information and target area of ​​the fruit tree to be sprayed based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data. The canopy structure information includes the canopy diameter, canopy height and canopy porosity. The spraying site determination module is configured to input the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters into a spraying operation control model that has been trained to a convergent state in a two-dimensional coordinate system of the fruit tree canopy, so as to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by a mapping function from flight parameters to disturbance area feature parameters and a mapping function from disturbance area feature parameters to droplet deposition feature parameters. The disturbance region determination module is configured to control the agricultural drone to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. It uses the inter-frame difference method to extract the dynamic region of the canopy image and extracts the two-dimensional projection point cloud of the three-dimensional point cloud data onto the canopy layers. Based on the dynamic region and the spatial displacement changes of the two-dimensional projection point cloud, it determines the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy. The first spraying operation module is configured to gradually fine-tune the hovering position of the plant protection drone along the horizontal axis and the vertical axis. After each adjustment, the characteristic parameters of the disturbed area are re-detected until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to perform the first spraying operation on the target area. The second spraying operation module is configured to detect the droplet deposition characteristic parameters corresponding to the target area. If the droplet deposition characteristic parameters do not reach a preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is then invoked to perform reverse calculations based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

[0014] An electronic device provided for another purpose of this application includes a central processing unit and a memory, the central processing unit being configured to invoke and run a computer program stored in the memory to perform the steps of the large-canopy fruit tree spraying operation control optimization method described in this application.

[0015] A computer-readable storage medium is provided for another purpose of this application, which stores, in the form of computer-readable instructions, a computer program implemented according to the method for controlling and optimizing spraying operations on large-canopy fruit trees, which, when called by a computer, executes the steps included in the corresponding method.

[0016] Compared to existing technologies, this application addresses the limitations of current agricultural drone spraying systems for controlling the spread of large-canopy fruit trees. Existing systems have limited spray width and rotor airflow disturbance, making it difficult to achieve uniform coverage of the entire canopy surface and interior. This results in unsatisfactory control effects, insufficient droplet deposition, and uneven target coverage. This application offers the following benefits, including but not limited to: Firstly, this application marks the target area in the form of boundary coordinates in a unified two-dimensional coordinate system of the canopy, thereby achieving precise definition of the target spatial range. This provides a standardized and quantifiable coordinate benchmark for the subsequent calculation of spraying sites, completely changing the current situation of vague target positioning and rough range definition in traditional operations.

[0017] Secondly, this application inputs canopy porosity, target area boundary coordinates, environmental parameters, and initial spray system parameters into a pre-trained and converged spraying operation control model, enabling precise output of initial spraying sites. By fully considering the effects of canopy porosity differences, ambient wind speed and temperature on rotor airflow, and the role of spray pressure in droplet diffusion, the output initial sites maximize the coverage potential of the rotor airflow disturbance area on the target. This application significantly improves the scientific rigor and accuracy of initial spraying sites, shortens the adjustment cycle for subsequent dynamic positioning, and reduces operational trial-and-error costs.

[0018] Thirdly, existing technologies for adjusting the hovering position of UAVs are mostly one-time coarse adjustments, which are difficult to adapt to the complex airflow environment of the canopy and often result in the disturbance area deviating from the target. This application adopts a dynamic positioning strategy of gradual fine-tuning and real-time re-examination. At the same time, it relies on the inter-frame difference method to extract the dynamic region and the dual-source technology of 3D point cloud to verify the displacement to confirm the characteristic parameters of the disturbance area. During the positioning process, the UAV makes small fine adjustments along the horizontal and vertical axes. After each adjustment, multi-source data is collected again to verify the disturbance area until it coincides with the target area. This avoids airflow turbulence caused by large displacement and eliminates the interference of non-rotor airflow such as wind blowing branches and leaves through dual-source verification, ensuring the authenticity and target relevance of the disturbance area. This solves the industry pain point that the disturbance area is difficult to match the target area due to the complex canopy structure of large-canopy fruit trees. It enables the rotor airflow to act accurately on the target area, creating the key airflow conditions for droplets to penetrate the canopy and be deposited uniformly.

[0019] Fourth, existing agricultural drones for supplementary spraying mostly adopt a non-discriminatory repeated spraying mode, which not only wastes pesticides but also fails to solve the problem of droplet penetration difficulty in low porosity areas. In contrast, this application targets areas with insufficient deposition, determines the optimal disturbance area and spraying site through model inverse calculation, and adjusts the spraying system parameters differently according to the canopy porosity. Through a step-by-step parameter adjustment strategy, it can specifically adapt to the droplet penetration requirements of different porosity areas, realizing the transformation from extensive supplementary spraying to precise enhancement.

[0020] Fifth, the operational process constructed in this application forms a complete closed loop from perception, decision-making, execution, feedback to optimization. During the respraying stage, data on "flight parameters - disturbance area - deposition effect" are recorded simultaneously and continuously added to the model training set, enabling iterative optimization of the spraying operation control model. Compared to the shortcomings of existing technologies where models are fixed and difficult to adapt to complex orchard environments, the closed-loop optimization mechanism of this application allows the model to continuously learn operational data under different fruit tree varieties, different canopy structures, and different environmental conditions, gradually improving the accuracy of site prediction and parameter adjustment. This makes the entire operational system more adaptable to different scenarios, providing technical support for the large-scale and standardized implementation of plant protection operations for large-canopy fruit trees. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the optimization method for spraying large-canopy fruit trees in the embodiments of this application; Figure 2 This is a schematic diagram illustrating a scenario for optimizing the spraying operation control of large-canopy fruit trees in an embodiment of this application. Figure 3 This is a schematic diagram of the control and optimization device for spraying large-canopy fruit trees in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the computer device in the embodiments of this application. Detailed Implementation

[0022] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0023] Those skilled in the art will understand that although the various methods in this application are described based on the same concept and thus present commonality among them, they can be performed independently unless otherwise specified. Similarly, the various embodiments disclosed in this application are all based on the same inventive concept; therefore, concepts expressed in the same way, as well as concepts that are appropriately changed for convenience but are expressed differently, should be understood equivalently.

[0024] Unless otherwise expressly stated, the various embodiments disclosed in this application can be combined in a cross-cutting manner to flexibly construct new embodiments, as long as such combination does not depart from the inventive spirit of this application and can meet the needs of the prior art or solve a certain deficiency in the prior art. Those skilled in the art should be aware of such modifications.

[0025] Please see Figure 1 and Figure 2 In one embodiment of the method for optimizing the control of spraying operations on large-canopy fruit trees, this application includes: Step S10: Based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, determine the canopy structure information and target area of ​​the fruit tree to be sprayed, wherein the canopy structure information includes canopy diameter, canopy height and canopy porosity. The large-canopy fruit tree spraying operation control and optimization system in the terminal equipment can acquire images of the fruit tree canopy to be sprayed and its corresponding three-dimensional point cloud data. Based on the images of the fruit tree canopy and its corresponding three-dimensional point cloud data, the system can determine the canopy structure information and target area of ​​the fruit tree to be sprayed. The fruit trees to be sprayed include lychee trees, longan trees, citrus trees, and mango trees, etc. The canopy structure information includes canopy diameter, canopy height, and canopy porosity. Specifically, in the canopy remote sensing stage, images of the fruit tree canopy to be sprayed are acquired by an image acquisition device mounted on a drone, and the corresponding three-dimensional point cloud data of the fruit tree canopy is acquired by remote sensing equipment such as LiDAR, so as to comprehensively obtain the canopy structure information of the fruit tree to be sprayed, including canopy diameter, canopy height, and canopy porosity.

[0026] In some embodiments, after determining the canopy structure information and target area of ​​the fruit tree to be sprayed based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, the process includes: A two-dimensional coordinate system for the fruit tree canopy is constructed with a preset location point as the origin. The structural information of the fruit tree canopy is mapped to the two-dimensional coordinate system for modeling to determine the canopy plane area. The coordinate boundary of the target area is marked on the two-dimensional coordinate system.

[0027] Specifically, the preset location point can be Key sites within a planar region Establish a two-dimensional coordinate system for the canopy with the origin as the origin, and the canopy structure information is as follows: Where D is the canopy diameter, H is the canopy height, and K is the canopy porosity; the acquired canopy structure information is mapped to the canopy two-dimensional coordinate system to form a visualized canopy spatial model. During modeling, irrelevant point clouds such as ground and weeds need to be removed, and the effective point cloud of the fruit tree canopy is retained; then the fruit tree canopy is layered according to the canopy height interval and the porosity distribution is marked. Different gray values ​​or colors are used to distinguish the high penetration area, medium penetration area and low penetration area in the canopy two-dimensional coordinate system.

[0028] More specifically, in litchi orchards, plant protection drones equipped with multispectral cameras and LiDAR are used to conduct cruise flights over target litchi trees to obtain information on the canopy structure of individual litchi trees, including canopy diameter, canopy height, canopy porosity, and the two-dimensional coordinate range of fruit concentration areas and pest and disease-prone areas.

[0029] Using the top left corner of the image returned by the UAV visual sensor as the origin, a planar coordinate system for the litchi tree canopy is established, and the acquired remote sensing data is mapped into this planar coordinate system to form a two-dimensional coordinate model of the litchi tree canopy.

[0030] The porosity distribution of the canopy was labeled according to height. Areas with porosity higher than 60% were marked as high penetration areas (H area), 30% to 60% as medium penetration areas (M area), and less than 30% as low penetration areas (L area), and these areas were marked with different colors in a two-dimensional coordinate system. At the same time, the coordinates of the fruit concentration areas (high incidence areas of litchi fruit borers) and leaf areas prone to diseases and pests identified by the multispectral camera were superimposed on the two-dimensional coordinate system. This coordinate system serves as a unified coordinate benchmark for perturbation identification and reverse inference of the optimal perturbation area.

[0031] In a further embodiment, after completing the modeling of the canopy two-dimensional coordinate system, the three-dimensional point cloud data of the fruit tree canopy is preprocessed and fused. The three-dimensional point cloud data of the fruit tree canopy, the fruit tree canopy map and environmental parameters are synchronized by timestamp to ensure that the multi-source data at the same time and the same spatial location correspond one-to-one. A coordinate transformation algorithm is used to convert the pixel coordinates in the image into the actual spatial coordinates in the canopy two-dimensional coordinate system.

[0032] In a further embodiment, by combining the image acquisition device and the vegetation index and other data obtained from previous field verification, the fruit concentration area and the suspected pest and disease area are identified as the target area of ​​the fruit tree canopy, and their coordinate boundaries are accurately marked in a two-dimensional coordinate system.

[0033] In the stage of determining the spraying site, the spraying site of the drone is initially determined by combining the target area with the pre-built "flight parameters-disturbance area-droplet deposition" model. At the same time, the visual sensor mounted on the drone and the sensing equipment deployed in the fruit tree canopy need to be activated.

[0034] During the dynamic positioning of the drone, the visual sensor continuously acquires images of disturbances on the surface of the fruit tree canopy. To ensure the accuracy of subsequent analysis, Gaussian filtering is first used to preprocess the images, filtering out interference factors such as ambient light fluctuations and sensor noise. Then, the inter-frame difference method is used to process adjacent frames, generating a difference image by calculating the difference in pixel grayscale values. Finally, a threshold for the rate of change of grayscale values ​​is set, and pixel areas in the difference image whose grayscale change rate exceeds the threshold are identified as dynamic areas of branches and leaves swaying due to the rotor airflow, thus filtering out candidate disturbance areas.

[0035] The canopy images acquired by the visual sensor are preprocessed to determine the minimum bounding rectangle of the disturbed region and record its coordinate range in the two-dimensional coordinate system of the canopy.

[0036] To improve the reliability of disturbance identification, time-series data of canopy point clouds acquired by lidar are further used to verify candidate disturbance regions: if the spatial displacement or density change of the point cloud exceeds a preset threshold per unit time, the region is confirmed as a valid disturbance region. The center coordinates of the disturbance region formed by the interaction between the current rotor airflow and the canopy are extracted and compared with the center coordinates of the target region to calculate the deviation. .

[0037] Finally, the hovering position of the drone is gradually fine-tuned according to the principle of small adjustments. After each adjustment, the image is re-acquired and the deviation is calculated until the disturbance area and the target area are highly overlapped. Then the spraying can be started. At this time, the hovering position of the drone is the precise spraying position.

[0038] Based on this, for the flight parameters of agricultural drones, the data of "disturbed area characteristic parameters - droplet deposition characteristic parameters" under different flight parameter states are recorded simultaneously, and the correlation model of "flight parameters - disturbed area - droplet deposition" is optimized and iterated.

[0039] During the droplet enhancement stage, if insufficient deposition occurs, the location for droplet enhancement is determined as needed. Based on the model, the characteristic parameters of the optimal disturbance area are derived, thereby determining the optimal spraying location for the UAV.

[0040] Meanwhile, relevant operating parameters are adjusted according to porosity adaptation priority: if the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the high penetration zone, the drone's operating altitude is adjusted first; if the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the medium penetration zone, the spray pressure is increased on the basis of reducing the altitude; if the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the low penetration zone, in addition to reducing the altitude and increasing the pressure, a nozzle with a smaller atomization particle size is replaced to ensure that the droplets accurately penetrate to the target deposition position.

[0041] The data of "flight parameters - disturbance region feature parameters - droplet deposition feature parameters" before and after parameter adjustment are recorded in real time and added to the model training dataset to optimize the back-inference accuracy of "droplet deposition feature parameters - disturbance region feature parameters".

[0042] Step S20: In the two-dimensional coordinate system of the fruit tree canopy, the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters are input into the spraying operation control model that has been trained to a convergent state, so as to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by the mapping function from flight parameters to disturbance area feature parameters and the mapping function from disturbance area feature parameters to droplet deposition feature parameters. Based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, after determining the canopy structure information and target area of ​​the fruit tree to be sprayed, in the two-dimensional coordinate system of the canopy, the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters are input into the spraying operation control model that has been trained to convergence state, so as to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by the mapping function from flight parameters to disturbance area feature parameters and the mapping function from disturbance area feature parameters to droplet deposition feature parameters. In some embodiments, the flight parameters characterize parameters related to the flight state of the agricultural drone during spraying operations, used to control the spatial position and rotor power output of the agricultural drone, and include rotor speed, hovering altitude, horizontal axis coordinate, and horizontal axis coordinate; the boundary coordinates of the target area characterize the coordinates of the diagonal vertex of the minimum bounding rectangle of the target area in the canopy two-dimensional coordinate system; the canopy porosity is a parameter characterizing the density of branches and leaves inside the fruit tree canopy, wherein the canopy porosity includes the average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone, and the average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone respectively corresponds to the density of branches and leaves in different areas of the fruit tree canopy. The parameters are defined as follows: The environmental parameters represent the external environmental conditions related to the spraying operation, including ambient wind speed and ambient temperature; the spraying system parameters represent the working status parameters of the agricultural drone's spraying system, including spray pressure and atomized particle size; the disturbance area characteristic parameters represent the spatial attribute parameters of the branch and leaf disturbance area formed by the interaction between the rotor airflow and the fruit tree canopy, including the coordinates of the center point in the canopy two-dimensional coordinate system, the area of ​​the minimum circumscribed rectangle, and the angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system; the droplet deposition characteristic parameters represent the parameters of the droplet adhesion effect in the target area after spraying, including the droplet deposition density and droplet coverage in the target area.

[0043] In some embodiments, the step of training a spraying operation control model includes: Step S201: Obtain a sample dataset, wherein the sample dataset includes multiple sets of sample data, the sample data includes input data and output data, the input data includes flight parameters of the agricultural drone, canopy porosity, environmental parameters and spray system parameters; the output data includes disturbance area characteristic parameters and droplet deposition characteristic parameters; Step S202: Fix the parameters of the spraying system, adjust the flight parameters, canopy porosity and environmental parameters of the plant protection drone, collect the characteristic parameters of the disturbed area corresponding to different parameter combinations, and use a neural network regression algorithm to iteratively optimize the first mean square error between the predicted value and the actual value of the characteristic parameters of the disturbed area until the first mean square error reaches below a preset threshold, so as to determine the mapping function from the flight parameters to the characteristic parameters of the disturbed area. Step S203: Fix the flight parameters, adjust the disturbance region feature parameters, canopy porosity and spray system parameters, and focus on supplementing sample data in the low penetration area to collect droplet deposition feature parameters corresponding to different parameter combinations; use a neural network regression algorithm to add error weights to the preset loss function to the correspondence between the disturbance region feature parameters and the target region feature parameters, and perform iterative optimization with the goal of minimizing the second mean square error between the predicted and actual values ​​of the droplet deposition feature parameters until the second mean square error value reaches below a preset threshold, so as to determine the mapping function from the disturbance region feature parameters to the droplet deposition feature parameters; Step S204: Integrate the mapping function from the flight parameters to the characteristic parameters of the disturbance area and the mapping function from the characteristic parameters of the disturbance area to the characteristic parameters of droplet deposition to construct an end-to-end spraying operation control model; Step S205: Select a portion of sample data as a test set to verify the model accuracy. If the prediction deviation of the perturbation area feature parameters and droplet deposition feature parameters exceeds the preset threshold, supplement the corresponding sample data and retrain the mapping function until the spraying operation control model meets the preset accuracy.

[0044] Specifically, let the flight parameter set of the agricultural drone be... ,in, The rotor speed, Hovering height The horizontal axis (X-axis) coordinate. Let the horizontal vertical axis (Y-axis) be the coordinate; let the set of characteristic parameters of the effective disturbance region be... , The coordinates of the center point of the disturbance region are: The area of ​​the disturbance region. The angle between the main extension direction of the disturbance region and the horizontal axis (X-axis).

[0045] The mapping function from flight parameters to characteristic parameters of the disturbance region is expressed as: ,in, For mapping functions, The set of canopy porosities, where , , The average porosity of the high-penetration zone (H zone), medium-penetration zone (M zone), and low-penetration zone (L zone) are respectively. For a set of environmental parameters, where, For ambient wind speed, This represents the ambient temperature, used to compensate for the impact of environmental factors on the propagation of rotor airflow.

[0046] In some embodiments, the set of droplet deposition characteristic parameters is set as follows: ,in, In the two-dimensional coordinate system of the canopy Deposition density at location, For the fog droplet coverage, the mapping function from the characteristic parameters of the disturbed region to the characteristic parameters of fog droplet deposition is expressed as: , In the formula, g is the mapping function, which is obtained by training multiple sets of "disturbance region characteristic parameters - droplet deposition characteristic parameters" sample data (such as more than 500 sets). During the training process, it is important to focus on associating the correspondence between the center point of the disturbance region and the high-density deposition area. Here, represents the spray system parameters, i.e., the spray pressure; K represents the canopy porosity set.

[0047] By combining the mapping functions from flight parameters to disturbance region characteristic parameters and from disturbance region characteristic parameters to droplet deposition characteristic parameters, a spraying operation control model is constructed, wherein the spraying operation control model is expressed as: ; The core function of this model is to take into account the flight parameters P of the agricultural drone, the canopy porosity K, the environmental parameters E, and the spray pressure. This will output the droplet deposition characteristic parameters S under the corresponding working conditions, providing a quantitative calculation basis for subsequent UAV dynamic positioning and parameter optimization. Meanwhile, the spraying operation control model supports inverse calculations, given the target droplet deposition characteristic parameters. From this, the optimal perturbation region characteristic parameters that satisfy the target droplet deposition characteristic parameters can be derived. and the corresponding optimal flight parameters .

[0048] Furthermore, the training of the spraying operation control model in this application revolves around "data preparation - phased training of the mapping function - model integration and optimization," and the specific steps include: The data preparation phase mainly involves collecting multi-source data around the model's input and output parameters. Input parameters include the agricultural drone's flight parameters P, canopy porosity K, environmental parameters E, and spray pressure. The output parameters include the perturbation region characteristic parameter I and the droplet deposition characteristic parameter S. The sample size needs to reach more than 500 groups and cover a variety of large-canopy fruit trees, different porosity zones and environmental conditions. At the same time, the data is preprocessed to remove outliers, realize multi-source data synchronization through timestamps, and standardize the parameters to the [0,1] interval to eliminate the influence of dimensional differences.

[0049] In the phased training phase of the mapping function, the mapping function f from the flight parameters to the characteristic parameters of the disturbance region is trained first, with a fixed spray pressure. Only the flight parameter P, canopy porosity K (actively selecting canopy regions with different porosities for sample data), and environmental parameter E are adjusted. The characteristic parameters of the disturbed region corresponding to different parameter combinations are collected. A BP neural network regression algorithm is used to iteratively optimize the algorithm with the goal of minimizing the mean square error between the predicted and actual values ​​of the disturbed region characteristic parameters until the error drops below a preset threshold. Then, a mapping function g from the disturbed region characteristic parameters to the droplet deposition characteristic parameters is trained: the flight parameter P is fixed, while the disturbed region characteristic parameters I, canopy porosity K, and spray pressure are adjusted. The model focuses on supplementing sample data in the low-penetration zone, using the same BP neural network regression algorithm for training, and adding error weights to the loss function to ensure that the model prioritizes fitting this key correlation until the prediction error of the droplet deposition feature parameters meets the preset requirements.

[0050] In the model integration phase, first follow The mapping function f from the trained flight parameters to the feature parameters of the disturbed area is integrated with the mapping function g from the feature parameters of the disturbed area to the feature parameters of droplet deposition into an end-to-end positive correlation spraying operation control model, realizing the quantitative calculation of "flight parameters - droplet deposition feature parameters". Then, 20% of the dataset is used as a test set to verify the model accuracy. If the prediction deviation of the feature parameters of the disturbed area and the feature parameters of droplet deposition exceeds the preset threshold, the corresponding sample data is supplemented and the mapping function is retrained until the spraying operation control model meets the preset accuracy.

[0051] Step S30: Control the plant protection drone to fly to the initial spraying site and hover, continuously collect images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data, use the inter-frame difference method to extract the dynamic region of the canopy image, and extract the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy layer. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, determine the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy of the fruit trees; In the two-dimensional coordinate system of the fruit tree canopy, the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters are input into the spraying operation control model that has been trained to convergence. After determining the initial spraying site corresponding to the initial flight parameters of the plant protection drone, the plant protection drone is controlled to fly to the initial spraying site and hover, continuously collecting images of the fruit tree canopy and its corresponding three-dimensional point cloud data. The dynamic region of the fruit tree canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the fruit tree canopy layers is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy are determined. In some embodiments, the step of extracting the dynamic region of the fruit tree canopy image using the inter-frame difference method and extracting the two-dimensional projection point cloud of the three-dimensional point cloud data in the fruit tree canopy layers, and determining the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy based on the spatial displacement changes of the dynamic region and the two-dimensional projection point cloud, includes: Step S301: Control the plant protection drone that has flown to the initial spraying site and hovers to continuously collect images of the canopy of the fruit trees to be sprayed and three-dimensional point cloud data that are spatiotemporally synchronized with the canopy images of the fruit trees. Specifically, by controlling a hovering drone to collect spatiotemporally synchronized images of the fruit tree canopy and 3D point cloud data, two things are achieved: First, it ensures that the two types of data correspond to the same time and the same spatial location within the canopy, avoiding mismatches between the dynamic region and the point cloud verification due to temporal or spatial misalignment, thus guaranteeing the correlation of multi-source data from the source. Second, it focuses on collecting data in the initial spraying site hovering scenario, eliminating the interference of drone body vibration on image and point cloud quality during flight, ensuring that the data can truly reflect the interaction between the rotor airflow and the canopy, providing a high-quality data source for subsequent identification of disturbed areas.

[0052] Step S302: After preprocessing the continuously acquired fruit tree canopy images, the pixel grayscale value difference between two adjacent frames of the fruit tree canopy images is calculated using the inter-frame difference method to generate a difference image. The pixel areas in the difference image where the grayscale value change rate exceeds the preset change rate threshold are identified as the dynamic areas of the fruit tree canopy affected by the rotor airflow. After preprocessing the canopy image, dynamic regions are extracted using the inter-frame difference method. Preprocessing (such as Gaussian filtering) can remove image noise (such as light fluctuations and leaf reflections), avoid noise interference with grayscale value calculation, and improve the clarity of the difference image. The inter-frame difference method can quickly capture dynamic regions formed by the movement of branches and leaves driven by the rotor airflow by comparing the pixel grayscale changes of adjacent frames, which is more timely than single-frame image analysis. By filtering by grayscale value change rate threshold, small grayscale fluctuations (such as slight swaying of branches and leaves caused by a light breeze) can be excluded, accurately locking the regions significantly affected by the rotor airflow and reducing the interference of non-target dynamic regions on subsequent analysis.

[0053] Step S303: Based on the canopy height of the fruit tree to be sprayed, the three-dimensional point cloud data is processed into layers of the fruit tree canopy at preset height intervals, and the two-dimensional projection point cloud of each layer of three-dimensional point cloud data in the two-dimensional coordinate system of the canopy is extracted to obtain the two-dimensional projection point cloud corresponding to each layer. This approach, which uses layered processing of 3D point clouds based on canopy height and extraction of 2D projected point clouds, addresses the core challenges in analyzing large-canopy fruit trees. Firstly, for the 3D structure of fruit tree canopies—sparse at the top and bottom, dense in the middle—layered processing avoids the occlusion of the lower and middle regions caused by overlapping point clouds across the entire canopy, ensuring that disturbances at each height level can be identified, especially in the lower and middle parts of the canopy where droplet deposition is insufficient. Secondly, projecting the 3D point cloud onto the canopy's 2D coordinate system aligns it with the 2D coordinate system of dynamic image regions, providing a unified benchmark for subsequent spatial location matching and avoiding matching errors caused by differences in coordinate dimensions. Thirdly, layered projected point clouds more clearly reflect the point cloud distribution at different height levels, providing more refined local data for subsequent time-series analysis (such as spatial displacement and density changes), and improving the accuracy of candidate disturbance region identification.

[0054] Step S304: Perform time-series analysis on the two-dimensional projection point cloud of each layer, calculate the spatial displacement and point cloud density change of the two-dimensional projection point cloud per unit time; and determine the two-dimensional projection point cloud regions where the spatial displacement exceeds a preset displacement threshold or the point cloud density change exceeds a preset density change threshold as candidate disturbance regions. The core value of performing a two-dimensional temporal analysis of layered two-dimensional projected point clouds, combining spatial displacement and density change, lies in quantitative verification. Firstly, by calculating the spatial displacement of the point cloud per unit time, it can be determined whether branches and leaves undergo significant movement (not just minor swaying) due to rotor airflow; displacement exceeding a threshold is marked as a potential disturbance area. Secondly, the density change of the point cloud reflects the dynamic changes in the canopy pores behind the branches and leaves driven by airflow (such as a decrease in local point cloud density due to branch separation), further corroborating the existence of disturbance. This dual-threshold screening is more rigorous than single-dimensional judgment, effectively eliminating false dynamics caused by non-rotor airflow factors (such as sudden gusts or insects flying by), ensuring that candidate disturbance areas are directly related to rotor airflow.

[0055] Step S305: Spatially match the dynamic region with the candidate disturbance regions of each layer, and select the candidate disturbance regions that spatially overlap with the dynamic region as the effective disturbance regions formed by the interaction between the rotor airflow and the fruit tree canopy; extract the center point coordinates, minimum circumscribed rectangle area, and angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system of the effective disturbance region to construct the characteristic parameters of the disturbance region.

[0056] By spatial matching and feature parameter extraction of dynamic regions and candidate disturbance regions, a dual guarantee of "qualitative + quantitative" is achieved: First, the spatial overlap screening of dynamic regions in the image and candidate regions in the point cloud completely eliminates misjudgments from a single data source (such as dynamic regions misjudged in the image without point cloud displacement support), ensuring the authenticity and accuracy of the final effective disturbance region; Second, feature parameters are constructed by extracting "center point coordinates, minimum bounding rectangle area, and angle of the main extension direction", which fully matches the quantitative requirements of the spraying operation control model for the disturbance region. This provides a positioning basis for subsequent UAV dynamic positioning (such as calculating deviation based on center point coordinates) and provides standardized input for the model to predict droplet deposition feature parameters, directly supporting the implementation of precise target spraying operations.

[0057] Step S40: Gradually fine-tune the hovering position of the agricultural drone along the horizontal axis and the vertical axis. After each adjustment, re-detect the characteristic parameters of the disturbance area until the disturbance area corresponding to the characteristic parameters of the disturbance area coincides with the target area, so as to perform the first spraying operation on the target area. The plant protection drone is controlled to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy layer is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy are determined. The hovering position of the plant protection drone is gradually fine-tuned along the horizontal axis and the vertical axis. After each adjustment, the characteristic parameters of the disturbance region are re-detected until the disturbance region corresponding to the characteristic parameters of the disturbance region coincides with the target region, so as to perform the first spraying operation on the target region. In some embodiments, the hovering position of the agricultural drone is gradually fine-tuned along the horizontal and vertical axes, and the characteristic parameters of the disturbed area are re-detected after each adjustment until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to perform the first spraying operation on the target area, including: Step S401: Based on the canopy two-dimensional coordinate system, compare the spatial position of the disturbance region corresponding to the disturbance region feature parameters with that of the target region, and determine whether the center point of the disturbance region coincides with the center point of the target region; if they do not coincide, calculate the first deviation value of the center point of the disturbance region relative to the center point of the target region in the horizontal axis direction and the second deviation value in the vertical axis direction of the canopy two-dimensional coordinate system. Step S402: Based on the first deviation value and the second deviation value, the hovering position of the agricultural drone is gradually fine-tuned along the horizontal axis and the vertical axis by a preset step size, wherein the preset fine-tuning step size is less than or equal to 0.1m. Step S403: After each adjustment of the hovering position, re-acquire the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, and re-execute the step of extracting the dynamic region of the canopy image of the fruit tree using the inter-frame difference method, and extracting the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy layer of the fruit tree, and determining the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy of the fruit tree based on the dynamic region combined with the spatial displacement change of the two-dimensional projection point cloud. Step S404: Repeat steps S401 to S403 until the overlap between the perturbation region corresponding to the re-detected perturbation region feature parameters and the target region in the canopy two-dimensional coordinate system reaches a preset percentage, and the area of ​​the perturbation region is greater than or equal to a certain multiple of the target region area; maintain the current hovering position, start the spraying system of the plant protection drone, and perform the first spraying operation on the target region according to the initially set spraying pressure and atomized particle size parameters, wherein the preset percentage is greater than or equal to 95%, and the certain multiple is greater than the value 1.

[0058] As can be seen from steps S401 to S404 above, by comparing the center point to quantify the horizontal and vertical axis deviation values, the fuzzy requirement of regional overlap is transformed into specific data, completely avoiding the blindness of experience-based positioning and providing a clear direction for subsequent fine-tuning, thus improving target-alignment efficiency from the source. Using small step adjustments of ≤0.1m avoids large displacements that could cause rotor airflow turbulence, ensuring the stability and controllability of the disturbed area, while also reducing the risk of the drone colliding with the canopy. This is particularly suitable for operational scenarios with large canopies and dense branches of fruit trees, balancing positioning accuracy and operational safety. The disturbed area is re-examined in real time using fruit tree canopy images and their corresponding 3D point cloud data, dynamically capturing the impact of positional changes on airflow, verifying whether the fine-tuning pushes the disturbed area closer to the target, and preventing the recurrence of invalid positioning. The repositioning is terminated based on two criteria: the overlap between the disturbed area and the target area in the canopy's two-dimensional coordinate system reaches a preset percentage, and the area of ​​the disturbed area is greater than or equal to a certain multiple of the target area. This dual standard ensures precise alignment between the disturbed area and the target while also covering the target edge with sufficient area to prevent missed spraying. After meeting these criteria, spraying is performed according to the initial parameters, balancing target relevance and operational efficiency, providing a preliminary guarantee for achieving the desired droplet deposition effect.

[0059] Step S50: Detect the droplet deposition characteristic parameters corresponding to the target area. If the droplet deposition characteristic parameters do not reach the preset threshold, the target area is determined as a target area with insufficient deposition. The spraying operation control model is called to perform reverse calculation based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

[0060] The hovering position of the agricultural drone is gradually fine-tuned along the horizontal and vertical axes. After each adjustment, the characteristic parameters of the disturbed area are re-detected until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area. After the first spraying operation is performed on the target area, the droplet deposition characteristic parameters corresponding to the target area are detected. If the droplet deposition characteristic parameters do not reach the preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is called to perform reverse calculation based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbed area characteristic parameters and the corresponding optimal spraying position. The spraying system parameters of the agricultural drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

[0061] In some embodiments, the step of calling the spraying operation control model to perform reverse calculations based on the boundary coordinates of the insufficient deposition target area to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site, and adjusting the spraying system parameters of the agricultural drone based on the canopy porosity corresponding to the insufficient deposition target area to perform a second spraying operation on the insufficient deposition target area until the droplet deposition characteristic parameters reach a preset threshold includes: Step S501: If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the high penetration area, then the operating height of the agricultural drone will be reduced by 5% to 10%. For high-penetration areas, the operating height is reduced by only 5%-10%, which improves deposition efficiency by shortening the droplet transmission distance. No additional adjustment of spray pressure or nozzle is required. While ensuring the effect of respraying, excessive consumption of energy and chemicals is avoided, thus balancing operating costs and spraying quality.

[0062] Step S502: If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the medium penetration zone, while reducing the operating height of the agricultural drone by 5% to 10%, the spray pressure of the spraying system will be increased by 10% to 15%. In the medium penetration zone, the spray pressure is increased by 10%-15% while reducing the operating height of agricultural drones by 5% to 10%. This shortens the transmission path and enhances the initial kinetic energy of the droplets, enabling them to effectively penetrate the medium-dense canopy. This addresses the pain points of "insufficient penetration and uneven deposition" in this area, achieving both droplet coverage and density targets.

[0063] Step S503: If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the low penetration area, on the basis of reducing the operating height of the plant protection drone by 5% to 10% and increasing the spray pressure of the spray system by 10% to 15%, the atomized particle size of the nozzle of the spray system will be reduced by 20% to 30%.

[0064] By overlaying low-penetration zones, the operating height of agricultural drones is reduced by 5% to 10%, the spray pressure of the spraying system is increased by 10% to 15%, and the atomized particle size of the spraying system nozzles is reduced by 20% to 30%. Through a combination of close-range spraying, strong initial kinetic energy, and small-diameter droplets, droplets can break through the dense canopy barrier and accurately reach the target area, completely solving the industry pain point of insufficient droplet deposition in low-porosity areas and ensuring that the droplet deposition characteristic parameters after respraying are stable and meet the standards.

[0065] As can be seen from the above embodiments, compared with the prior art, the present application addresses the problem that existing plant protection drone spraying systems have limited spray width and rotor airflow disturbance for the control of large-canopy fruit trees, making it difficult to achieve uniform coverage of the entire canopy surface and interior, resulting in unsatisfactory control effects, insufficient droplet deposition, and uneven target coverage. The present application has, but is not limited to, the following beneficial effects: Firstly, this application marks the target area in the form of boundary coordinates in a unified two-dimensional coordinate system of the canopy, thereby achieving precise definition of the target spatial range. This provides a standardized and quantifiable coordinate benchmark for the subsequent calculation of spraying sites, completely changing the current situation of vague target positioning and rough range definition in traditional operations.

[0066] Secondly, this application inputs canopy porosity, target area boundary coordinates, environmental parameters, and initial spray system parameters into a pre-trained and converged spraying operation control model, enabling precise output of initial spraying sites. By fully considering the effects of canopy porosity differences, ambient wind speed and temperature on rotor airflow, and the role of spray pressure in droplet diffusion, the output initial sites maximize the coverage potential of the rotor airflow disturbance area on the target. This application significantly improves the scientific rigor and accuracy of initial spraying sites, shortens the adjustment cycle for subsequent dynamic positioning, and reduces operational trial-and-error costs.

[0067] Thirdly, existing technologies for adjusting the hovering position of UAVs are mostly one-time coarse adjustments, which are difficult to adapt to the complex airflow environment of the canopy and often result in the disturbance area deviating from the target. This application adopts a dynamic positioning strategy of gradual fine-tuning and real-time re-examination. At the same time, it relies on the inter-frame difference method to extract the dynamic region and the dual-source technology of 3D point cloud to verify the displacement to confirm the characteristic parameters of the disturbance area. During the positioning process, the UAV makes small fine adjustments along the horizontal and vertical axes. After each adjustment, multi-source data is collected again to verify the disturbance area until it coincides with the target area. This avoids airflow turbulence caused by large displacement and eliminates the interference of non-rotor airflow such as wind blowing branches and leaves through dual-source verification, ensuring the authenticity and target relevance of the disturbance area. This solves the industry pain point that the disturbance area is difficult to match the target area due to the complex canopy structure of large-canopy fruit trees. It enables the rotor airflow to act accurately on the target area, creating the key airflow conditions for droplets to penetrate the canopy and be deposited uniformly.

[0068] Fourth, existing agricultural drones for supplementary spraying mostly adopt a non-discriminatory repeated spraying mode, which not only wastes pesticides but also fails to solve the problem of droplet penetration difficulty in low porosity areas. In contrast, this application targets areas with insufficient deposition, determines the optimal disturbance area and spraying site through model inverse calculation, and adjusts the spraying system parameters differently according to the canopy porosity. Through a step-by-step parameter adjustment strategy, it can specifically adapt to the droplet penetration requirements of different porosity areas, realizing the transformation from extensive supplementary spraying to precise enhancement.

[0069] Fifth, the operational process constructed in this application forms a complete closed loop from perception, decision-making, execution, feedback to optimization. During the respraying stage, data on "flight parameters - disturbance area - deposition effect" are recorded simultaneously and continuously added to the model training set, enabling iterative optimization of the spraying operation control model. Compared to the shortcomings of existing technologies where models are fixed and difficult to adapt to complex orchard environments, the closed-loop optimization mechanism of this application allows the model to continuously learn operational data under different fruit tree varieties, different canopy structures, and different environmental conditions, gradually improving the accuracy of site prediction and parameter adjustment. This makes the entire operational system more adaptable to different scenarios, providing technical support for the large-scale and standardized implementation of plant protection operations for large-canopy fruit trees.

[0070] Please see Figure 3This application provides a large-canopy fruit tree spraying operation control optimization device, comprising a canopy image acquisition module 1100, a spraying site determination module 1200, a disturbance area determination module 1300, a first spraying operation module 1400, and a second spraying operation module 1500. The canopy image acquisition module 1100 is configured to determine the canopy structure information and target area of ​​the fruit tree to be sprayed based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data. The canopy structure information includes canopy diameter, canopy height, and canopy porosity. The spraying site determination module 1200 is configured to input the canopy porosity, the boundary coordinates of the target area, current environmental parameters, and initial spraying system parameters into a spraying operation control model trained to convergence in a two-dimensional coordinate system of the fruit tree canopy to determine the plant protection operation control model. The initial spraying site corresponds to the initial flight parameters of the drone. The spraying operation control model is constructed from a mapping function of flight parameters to disturbance area feature parameters and a mapping function of disturbance area feature parameters to droplet deposition feature parameters. The disturbance area determination module 1300 is configured to control the agricultural drone to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data onto the canopy layers is extracted. Based on the spatial displacement changes of the dynamic region combined with the two-dimensional projected point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy are determined; the first spraying operation module 1400 is configured to gradually fine-tune the hovering position of the plant protection drone along the horizontal and vertical axes, and re-detect the characteristic parameters of the disturbance region after each adjustment until the disturbance region corresponding to the characteristic parameters of the disturbance region coincides with the target region, so as to perform the first spraying operation on the target region; the second spraying operation module 1500 is configured to detect the droplet deposition corresponding to the target region. If the droplet deposition characteristic parameters do not reach a preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is then invoked to perform reverse calculations based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

[0071] Based on any embodiment of this application, please refer to Figure 4 Another embodiment of this application also provides an electronic device, which can be implemented by a computer device, such as... Figure 4 The diagram shows the internal structure of a computer device. This computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable storage medium stores an operating system, a database, and computer-readable instructions. The database stores control information sequences. When the processor executes the computer-readable instructions, it enables the processor to implement a method for optimizing the spraying operation of large-canopy fruit trees. The processor provides computing and control capabilities to support the operation of the entire computer device. The memory stores computer-readable instructions, which, when executed by the processor, enable the processor to execute the method for optimizing the spraying operation of large-canopy fruit trees as described in this application. The network interface of the computer device is used for communication with a terminal. Those skilled in the art will understand that… Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0072] In this embodiment, the processor is used to execute... Figure 3 The specific functions of each module are defined within the device, and the memory stores the program code and various data required to execute these modules. A network interface is used for data transmission between the user terminal and the server. In this embodiment, the memory stores the program code and data required to execute all modules in the large-canopy fruit tree spraying operation control and optimization device of this application. The server can call the server's program code and data to execute the functions of all modules.

[0073] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the large-canopy fruit tree spraying operation control optimization method described in any embodiment of this application.

[0074] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the large-canopy fruit tree spraying operation control optimization method described in any embodiment of this application.

[0075] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0076] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for optimizing the control of spraying operations on large-canopy fruit trees, characterized in that, include: Based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data, the canopy structure information and target area of ​​the fruit tree to be sprayed are determined. The canopy structure information includes canopy diameter, canopy height and canopy porosity. In the two-dimensional coordinate system of the fruit tree canopy, the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters are input into the spraying operation control model that has been trained to a convergent state to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by a mapping function from flight parameters to disturbance area feature parameters and a mapping function from disturbance area feature parameters to droplet deposition feature parameters. The agricultural drone is controlled to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy layer is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy are determined. The hovering position of the agricultural drone is gradually fine-tuned along the horizontal and vertical axes. After each adjustment, the characteristic parameters of the disturbed area are re-detected until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to carry out the first spraying operation on the target area. The droplet deposition characteristic parameters corresponding to the target area are detected. If the droplet deposition characteristic parameters do not reach a preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is called to perform reverse calculation based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

2. The method for optimizing spraying operations on large-canopy fruit trees according to claim 1, characterized in that, The flight parameters characterize the flight status parameters of the agricultural drone during spraying operations, and are used to control the spatial position and rotor power output of the agricultural drone. They include rotor speed, hovering height, horizontal axis coordinate, and horizontal axis coordinate. The boundary coordinates of the target region represent the coordinates of the diagonal vertices of the smallest bounding rectangle of the target region in the canopy two-dimensional coordinate system; The canopy porosity is a parameter that characterizes the density of branches and leaves inside the canopy of a fruit tree. The canopy porosity includes the average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone. The average porosity of the high-penetration zone, the medium-penetration zone, and the low-penetration zone correspond to the proportion of gaps between branches and leaves in different areas of the fruit tree canopy. The environmental parameters characterize the external environmental conditions related to the spraying operation, including ambient wind speed and ambient temperature. The spray system parameters characterize the working status parameters of the spray system of the agricultural drone, including spray pressure and atomized particle size; The disturbance region characteristic parameters characterize the spatial attribute parameters of the branch and leaf disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy. These parameters include the coordinates of the center point in the canopy two-dimensional coordinate system, the area of ​​the minimum circumscribed rectangle, and the angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system. The droplet deposition characteristic parameters characterize the adhesion effect of droplets in the target area after spraying, including the droplet deposition density and droplet coverage in the target area.

3. The method for optimizing spraying operations on large-canopy fruit trees according to claim 2, characterized in that, The steps for training a spraying operation control model include: Obtain a sample dataset, wherein the sample dataset includes multiple sets of sample data, the sample data includes input data and output data, the input data includes flight parameters of the agricultural drone, canopy porosity, environmental parameters and spray system parameters; the output data includes disturbance area characteristic parameters and droplet deposition characteristic parameters; The parameters of the spraying system are fixed, and the flight parameters, canopy porosity and environmental parameters of the agricultural drone are adjusted to collect the characteristic parameters of the disturbed area corresponding to different parameter combinations. A neural network regression algorithm is used to iteratively optimize the data with the goal of minimizing the first mean square error between the predicted and actual values ​​of the characteristic parameters of the disturbed area until the first mean square error reaches below a preset threshold, so as to determine the mapping function from the flight parameters to the characteristic parameters of the disturbed area. The flight parameters are fixed, and the characteristic parameters of the disturbed area, the canopy porosity, and the spray system parameters are adjusted. The sample data of the low penetration area is supplemented in particular to collect the droplet deposition characteristic parameters corresponding to different parameter combinations. A neural network regression algorithm is used to add the error weight of the correspondence between the characteristic parameters of the disturbed area and the characteristic parameters of the target area to the preset loss function. The algorithm is iteratively optimized with the goal of minimizing the second mean square error between the predicted and actual values ​​of the droplet deposition characteristic parameters until the second mean square error value reaches below the preset threshold, so as to determine the mapping function from the characteristic parameters of the disturbed area to the droplet deposition characteristic parameters. The mapping functions from flight parameters to disturbance area characteristic parameters and from disturbance area characteristic parameters to droplet deposition characteristic parameters are integrated to construct an end-to-end spraying operation control model. Select a portion of sample data as a test set to verify the model accuracy. If the prediction deviation of the perturbation area feature parameters and droplet deposition feature parameters exceeds the preset threshold, supplement the corresponding sample data and retrain the mapping function until the spraying operation control model meets the preset accuracy.

4. The method for optimizing spraying operations on large-canopy fruit trees according to claim 1, characterized in that, After determining the canopy structure information and target area of ​​the fruit trees to be sprayed based on the canopy image of the fruit trees to be sprayed and its corresponding 3D point cloud data, the process includes: A two-dimensional coordinate system for the fruit tree canopy is constructed with a preset location point as the origin. The structural information of the fruit tree canopy is mapped to the two-dimensional coordinate system for modeling to determine the canopy plane area. The coordinate boundary of the target area is marked on the two-dimensional coordinate system.

5. The method for optimizing spraying operations on large-canopy fruit trees according to claim 1, characterized in that, The steps of extracting the dynamic region of the fruit tree canopy image using the inter-frame difference method and extracting the two-dimensional projection point cloud of the three-dimensional point cloud data onto the fruit tree canopy layers, and determining the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the fruit tree canopy based on the spatial displacement changes of the dynamic region and the two-dimensional projection point cloud, include: Control the plant protection drone that has flown to the initial spraying site and hovers to continuously collect images of the canopy of the fruit trees to be sprayed, as well as three-dimensional point cloud data that is spatiotemporally synchronized with the canopy images of the fruit trees. After preprocessing the continuously acquired fruit tree canopy images, the inter-frame difference method is used to calculate the pixel gray value difference between two adjacent frames of the fruit tree canopy images to generate a difference image. The pixel areas in the difference image where the gray value change rate exceeds the preset change rate threshold are identified as the dynamic areas of the fruit tree canopy affected by the rotor airflow. Based on the canopy height of the fruit trees to be sprayed, the three-dimensional point cloud data is processed into layers of the fruit tree canopy at preset height intervals. The two-dimensional projection point cloud of each layer of three-dimensional point cloud data in the two-dimensional coordinate system of the canopy is extracted to obtain the two-dimensional projection point cloud corresponding to each layer. A time series analysis is performed on the two-dimensional projected point cloud of each layer to calculate the spatial displacement and point cloud density change of the two-dimensional projected point cloud per unit time; the two-dimensional projected point cloud regions where the spatial displacement exceeds a preset displacement threshold or the point cloud density change exceeds a preset density change threshold are identified as candidate disturbance regions. The dynamic region is spatially matched with the candidate disturbance regions of each layer, and the candidate disturbance regions that spatially overlap with the dynamic region are selected as the effective disturbance regions formed by the interaction between the rotor airflow and the fruit tree canopy. The center point coordinates, the area of ​​the minimum circumscribed rectangle, and the angle between the main extension direction and the horizontal axis of the canopy two-dimensional coordinate system of the effective disturbance region are extracted to construct the characteristic parameters of the disturbance region.

6. The method for optimizing spraying operations on large-canopy fruit trees according to claim 2, characterized in that, The steps of gradually fine-tuning the hovering position of the agricultural drone along the horizontal and vertical axes, re-detecting the characteristic parameters of the disturbed area after each adjustment, until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, and then performing the first spraying operation on the target area, include: Based on the canopy two-dimensional coordinate system, the spatial position of the disturbance region corresponding to the disturbance region feature parameters is compared with that of the target region to determine whether the center point of the disturbance region coincides with the center point of the target region; if they do not coincide, the first deviation value of the center point of the disturbance region relative to the center point of the target region in the horizontal axis direction and the second deviation value in the vertical axis direction of the canopy two-dimensional coordinate system are calculated. Based on the first deviation value and the second deviation value, the hovering position of the agricultural drone is gradually fine-tuned along the horizontal axis and the vertical axis by a preset step size, wherein the preset fine-tuning step size is less than or equal to 0.1m; After each adjustment of the hovering position, the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data are re-acquired. The dynamic region of the canopy image is extracted using the inter-frame difference method, and the two-dimensional projection point cloud of the three-dimensional point cloud data in the canopy is extracted. Based on the dynamic region and the spatial displacement change of the two-dimensional projection point cloud, the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy are determined. Repeat the above steps until the overlap between the perturbation region corresponding to the re-detected perturbation region feature parameters and the target region in the canopy two-dimensional coordinate system reaches a preset percentage, and the area of ​​the perturbation region is greater than or equal to a certain multiple of the target region area; maintain the current hovering position, start the spraying system of the plant protection drone, and perform the first spraying operation on the target region according to the initially set spraying pressure and atomized particle size parameters, wherein the preset percentage is greater than or equal to 95%, and the certain multiple is greater than the value 1.

7. The method for optimizing spraying operations on large-canopy fruit trees according to any one of claims 1 to 6, characterized in that, The steps include: calling the spraying operation control model to perform reverse calculations based on the boundary coordinates of the insufficient deposition target area to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site; and adjusting the spraying system parameters of the agricultural drone based on the canopy porosity corresponding to the insufficient deposition target area to perform a second spraying operation on the insufficient deposition target area until the droplet deposition characteristic parameters reach a preset threshold. If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the high penetration area, the operating height of the agricultural drone will be reduced by 5% to 10%. If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the medium penetration zone, the spray pressure of the spraying system will be increased by 10% to 15%, while reducing the operating height of the agricultural drone by 5% to 10%. If the canopy porosity corresponding to the insufficient deposition target area reaches the average porosity of the low penetration area, in addition to reducing the operating height of the agricultural drone by 5% to 10% and increasing the spray pressure of the spray system by 10% to 15%, the atomized particle size of the nozzle of the spray system will be reduced by 20% to 30%.

8. A device for controlling and optimizing spraying operations on large-canopy fruit trees, characterized in that, include: The canopy image acquisition module is configured to determine the canopy structure information and target area of ​​the fruit tree to be sprayed based on the canopy image of the fruit tree to be sprayed and its corresponding three-dimensional point cloud data. The canopy structure information includes the canopy diameter, canopy height and canopy porosity. The spraying site determination module is configured to input the canopy porosity, the boundary coordinates of the target area, the current environmental parameters, and the initial spraying system parameters into a spraying operation control model that has been trained to a convergent state in a two-dimensional coordinate system of the fruit tree canopy, so as to determine the initial spraying site corresponding to the initial flight parameters of the plant protection drone. The spraying operation control model is constructed by a mapping function from flight parameters to disturbance area feature parameters and a mapping function from disturbance area feature parameters to droplet deposition feature parameters. The disturbance region determination module is configured to control the agricultural drone to fly to the initial spraying site and hover, continuously acquiring images of the canopy of the fruit trees to be sprayed and their corresponding three-dimensional point cloud data. It uses the inter-frame difference method to extract the dynamic region of the canopy image and extracts the two-dimensional projection point cloud of the three-dimensional point cloud data onto the canopy layers. Based on the dynamic region and the spatial displacement changes of the two-dimensional projection point cloud, it determines the characteristic parameters of the disturbance region formed by the interaction between the rotor airflow and the canopy. The first spraying operation module is configured to gradually fine-tune the hovering position of the plant protection drone along the horizontal axis and the vertical axis. After each adjustment, the characteristic parameters of the disturbed area are re-detected until the disturbed area corresponding to the characteristic parameters of the disturbed area coincides with the target area, so as to perform the first spraying operation on the target area. The second spraying operation module is configured to detect the droplet deposition characteristic parameters corresponding to the target area. If the droplet deposition characteristic parameters do not reach a preset threshold, the target area is identified as a target area with insufficient deposition. The spraying operation control model is then invoked to perform reverse calculations based on the boundary coordinates of the target area with insufficient deposition to determine the optimal disturbance area characteristic parameters and the corresponding optimal spraying site. The spraying system parameters of the plant protection drone are adjusted according to the canopy porosity corresponding to the target area with insufficient deposition to perform a second spraying operation on the target area with insufficient deposition until the droplet deposition characteristic parameters reach the preset threshold, thereby completing the spraying operation control optimization for large-canopy fruit trees.

9. An electronic device comprising a central processing unit and a memory, characterized in that, The central processing unit is used to invoke and run a computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 7, which, when invoked by a computer, executes the steps included in the corresponding method.

Images