Unmanned aerial vehicle prevention and control method and system for monitoring diseases and insect pests

Abstract

The invention provides an unmanned aerial vehicle prevention and control method and system for disease and insect pest monitoring, and relates to the technical field of path planning, and the method comprises the steps: controlling an unmanned aerial vehicle to fly along an initial inspection path; determining disease and pest occurrence points, and determining a disease and pest occurrence point set according to the disease and pest occurrence points; determining a plurality of risk areas; determining a target hovering entry point based on each risk area; the current position and environment information of the unmanned aerial vehicle are obtained, a three-dimensional discrete grid field is constructed based on the target hovering entry point and the current position and environment information of the unmanned aerial vehicle, and the three-dimensional discrete grid field comprises a plurality of grid units and corresponding grid risk values; and in the three-dimensional discrete grid field, generating a plurality of candidate tracks based on an RRT algorithm, and determining a transition track based on a target hovering entry point and a grid risk value in combination with wind field analysis. The selection of the target hovering entry point effectively avoids the problem of unstable flight caused by local airflow disturbance in the approaching and hovering process of the unmanned aerial vehicle.

Description

Technical Field

[0001] This application relates to the field of path planning technology, and in particular to a method and system for the prevention and control of pests and diseases using unmanned aerial vehicles (UAVs). Background Technology

[0002] In existing agricultural drone pest and disease control operations, pre-set flight paths are typically used for uniform spraying, or the operational boundaries are manually demarcated after identifying the affected area based on remote sensing images. The drone is then controlled to fly over the target area and hover to perform precise spraying. To achieve precise application, the drone needs to approach the high-incidence area through a specific "entry point," which is usually selected by the operator based on experience, typically the geometric center of the affected area or the most visually open location. However, existing technologies often employ simple straight-line flight or circular detours to plan the transition trajectory from the current flight path to this entry point, lacking a comprehensive consideration of environmental risks and flight stability.

[0003] However, the selection of entry points lacks scientific basis and often fails to consider factors such as local airflow disturbances, the impact of obstacle wakes, and canopy turbulence. This makes drones susceptible to wind interference and unstable flight during approach and hovering. At the same time, the planned trajectories are mostly geometric shortest paths, failing to be optimized in combination with the distribution density of the disease, ignoring the priority coverage needs of high-incidence areas, and unable to be dynamically adjusted according to real-time wind speed and direction. This makes it difficult to effectively avoid dynamic risk areas, affecting operational safety and spraying accuracy. Summary of the Invention

[0004] This application aims to at least partially address one of the technical problems in the related art.

[0005] To achieve the above objectives, this application proposes a drone-based pest control method for monitoring pests and diseases, comprising the following steps:

[0006] Step 1: Obtain the initial inspection path, control the drone to fly along the initial inspection path, and collect crop canopy image sequences in real time;

[0007] Step 2: Determine the pest and disease occurrence points based on the canopy image sequence, and determine the pest and disease occurrence point set based on the pest and disease occurrence points;

[0008] Step 3: Based on the set of pest and disease occurrence points, determine several risk areas; determine the target hovering entry point based on each risk area;

[0009] Step 4: Obtain the current position and environmental information of the UAV. Based on the target hovering entry point, the current position and environmental information of the UAV, construct a three-dimensional discrete grid field. The three-dimensional discrete grid field includes several grid cells and corresponding grid risk values.

[0010] Step 5: Within the three-dimensional discrete grid field, multiple candidate trajectories are generated based on the RRT algorithm, and the transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis.

[0011] Further, determining the target hovering entry point based on each of the aforementioned risk areas includes the following steps:

[0012] Step 31: Based on the set of pest and disease occurrence points in the risk area, kernel density estimation is performed to obtain the density of pest and disease occurrence points, and the set of density peak points is determined based on the density of pest and disease occurrence points.

[0013] Step 32: Determine the geometric center based on the risk area, and construct a guide ray based on the nearest dense peak point of the geometric center;

[0014] Step 33: Based on the guiding ray, construct multiple candidate anchor points;

[0015] Step 34: Preset the spraying operation height; determine the crop canopy height map based on the crop canopy image sequence, and determine the average crop canopy height below each candidate anchor based on the crop canopy height map; determine the height difference based on the spraying operation height and the average crop canopy height, and determine whether the height difference is greater than or equal to 0.3m and less than or equal to 0.5m. If yes, proceed to the next step; if no, end.

[0016] Step 35: Obtain a 3D point cloud map of the work area, and extract the spatial envelope of static obstacles based on the 3D point cloud map;

[0017] Step 36: Based on the candidate anchor points, project them vertically onto the ground to obtain horizontal projection points; construct a search space with the horizontal projection points as the center. If the search space does not contain a static obstacle space envelope, the process ends; if the search space contains one or more static obstacle space envelopes, proceed to the next step.

[0018] Step 37: Real-time acquisition of wind speed and direction; determination of the prevailing wind direction vector based on the wind direction; determination of the first obstacle based on static obstacles; formation of a wake prediction zone by extending along the opposite direction of the prevailing wind direction vector to the leeward side based on the first obstacle; termination if the horizontal projection point is located in the wake prediction zone; termination if the horizontal projection point is not located in the wake prediction zone, using the current candidate anchor point as the target hovering entry point.

[0019] Furthermore, if multiple horizontal projection points are located within the wake prediction region, the first verified candidate anchor point will be used as the target hovering entry point.

[0020] Furthermore, a three-dimensional discrete grid field is constructed based on the target hovering entry point, the current position of the UAV, and environmental information, including:

[0021] Step 41: Divide the work area into a three-dimensional discrete grid; each grid cell of the three-dimensional discrete grid includes a first coordinate and a resolution;

[0022] Step 42: Initialize the grid risk value to 0 for each grid cell;

[0023] Step 43: Traverse the spatial envelope of all static obstacles. If an obstacle intersects with the current grid cell, mark the grid risk value of the current grid cell as the maximum risk value and mark the current grid cell as impassable.

[0024] Step 44: Based on the grid cells without marked impassable states and the vertical coordinate Z of the current grid cell satisfying: hc(X,Y)+0.3≤Z≤hc(X,Y)+0.8, it is determined that the current grid cell is located in the turbulent influence layer above the canopy; if not, the grid risk value remains unchanged; for the turbulent influence layer above the canopy, the canopy density of the grid cells without marked impassable states is determined based on the crop canopy height map, the perturbation value is determined based on the canopy density, and the first risk value is obtained based on the grid risk value and the perturbation value; where hc(X,Y) is the absolute elevation of the crop canopy height map at the horizontal position (X,Y);

[0025] Step 45: If, based on grid cells that are not marked as impassable, the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is greater than the target threshold, the current grid cell is determined to be located in a risk depression; if the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is less than or equal to the target threshold, the grid risk value remains unchanged; for risk depressions, a guiding value is determined based on the density of pest and disease occurrence points, and a second risk value is obtained based on the guiding value and the grid risk value;

[0026] Step 46: For grid cells located in the wake prediction area that are not marked as impassable, determine the wind-induced disturbance volume based on the real-time wind speed, and obtain the third risk value based on the wind-induced disturbance volume and the grid risk value.

[0027] Step 47: Based on all grid cells and their corresponding first coordinates, resolution, grid risk value or maximum risk value or first risk value or second risk value or third risk value, a three-dimensional discrete grid field is constructed.

[0028] Furthermore, if the second risk value is less than 0, then the second risk value is marked as 0.

[0029] Furthermore, the transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis, including:

[0030] Step 51: Based on each candidate trajectory, extract the sequence of all raster cells passed through;

[0031] Step 52: Determine the target risk value based on the grid cell sequence;

[0032] Step 53: If the target risk value is greater than or equal to the preset safety threshold, the current candidate trajectory is eliminated; if the target risk value is less than the preset safety threshold, the current candidate trajectory is retained and marked as the expected trajectory.

[0033] Step 54: Preset target wind speed range and target wind direction range, wherein the target wind speed range includes several wind speeds and the target wind direction range includes several wind directions; based on the several wind speeds in the target wind speed range and the several wind directions in the target wind direction range, construct multiple wind field scenario combinations.

[0034] Step 55: Based on each wind field scenario combination and the real-time wind speed and direction, obtain the first wind speed and first wind direction for a preset time interval;

[0035] Step 56: Using the first wind speed and the first wind direction, generate the first wake prediction area for each wind field scenario combination and readjust the grid risk value of the three-dimensional discrete grid field.

[0036] Step 57: Based on the readjusted grid risk value, determine the risk exposure of the UAV flying under each wind field scenario combination, and based on the risk exposure, select transition trajectories from the expected trajectories.

[0037] Furthermore, based on risk exposure, transitional trajectories are screened from the expected trajectories, including:

[0038] Step 571: Determine whether the risk exposure is greater than the target exposure value;

[0039] Step 572: Locate the last grid cell whose expected trajectory is less than or equal to the safety threshold and mark it as the midpoint;

[0040] Step 573: Determine the distance between the midpoint and the expected trajectory endpoint, and mark it as the adjustment distance;

[0041] Step 574: Determine whether the adjustment distance is less than or equal to the minimum correction distance;

[0042] Step 575: If the judgment result of step 571 and / or step 574 is yes, then the current expected trajectory is removed; otherwise, the expected trajectory is defined as a transition trajectory.

[0043] This invention provides a drone-based pest control system for monitoring pests and diseases, comprising the following modules:

[0044] Inspection module: used to obtain the initial inspection path, control the UAV to fly along the initial inspection path, and collect crop canopy image sequences in real time;

[0045] Pest and disease occurrence point determination module: used to determine the occurrence points of pests and diseases based on the canopy image sequence, and to determine the set of occurrence points of pests and diseases based on the occurrence points of pests and diseases;

[0046] Target hovering entry point construction module: used to determine several risk areas based on the set of pest and disease occurrence points; and to determine a target hovering entry point based on each risk area.

[0047] 3D Discrete Grid Field Construction Module: Used to acquire the current position and environmental information of the UAV, and construct a 3D discrete grid field based on the target hovering entry point, the current position and environmental information of the UAV. The 3D discrete grid field includes several grid cells and corresponding grid risk values.

[0048] Transition trajectory construction module: used to generate multiple candidate trajectories based on the RRT algorithm within the three-dimensional discrete grid field, and determine the transition trajectory based on the target hovering entry point and grid risk value, combined with wind field analysis.

[0049] Compared with existing technologies, this application provides a drone-based pest control method for monitoring pests and diseases. By constructing a three-dimensional discrete grid field, the method comprehensively considers the spatial envelope of static obstacles, the turbulent influence layer above the canopy, the risk depressions formed by the density of pest and disease occurrence points, and the wake prediction zone dynamically generated based on real-time wind speed and direction during path planning, achieving refined modeling of the operating environment. This allows the selection of the target hovering entry point to no longer rely on empirical settings, but rather on scientific optimization based on multiple parameters such as airflow stability, obstacle wake influence, canopy turbulence intensity, and pest distribution density, effectively avoiding flight instability caused by local airflow disturbances during drone approach and hovering. Simultaneously, the trajectory generation process overcomes the limitations of traditional geometric shortest paths. By introducing a pest distribution guidance mechanism, it prioritizes coverage of high-incidence areas, improving the targeting and coverage of spraying operations. Furthermore, by dynamically updating the wake prediction zone and grid risk value, the path possesses real-time wind field adaptability, enabling proactive avoidance of dynamic risk areas caused by wind field changes. Based on this, the safety of drone flight, operational stability, and application accuracy are significantly improved.

[0050] This application achieves a breakthrough improvement over the traditional empirical and fixed entry point selection method by setting up an automatic selection mechanism for target hovering entry points based on kernel density estimation of pest and disease occurrence points and multi-constraint collaborative screening. Specifically, firstly, the densely populated disease areas within the risk zone are accurately identified through kernel density estimation, and the set of density peak points is determined. A guiding ray is constructed from the geometric center of the region to the nearest density peak point to ensure that the path planning has an operation priority orientation towards high-incidence areas. On this basis, multiple candidate anchor points are generated. Combined with the matching analysis of the spraying operation height and the actual height of the crop canopy, a reasonable hovering height window with a height difference of 0.3m to 0.5m is selected to ensure the uniformity of droplet deposition and the requirement for anti-drift. Furthermore, the spatial envelope of static obstacles is extracted by integrating a 3D point cloud map to eliminate candidate points with structural obstruction. Finally, real-time wind field perception is introduced to construct a wake prediction zone. High-risk projection points located within the leeward vortex influence range of obstacles are eliminated, and only candidate anchor points in the stable airflow zone are retained as the final target hovering entry points. It significantly improves the flight stability and hovering safety of drones when approaching targets, effectively avoids problems such as pesticide drift, uneven coverage or flight loss of control caused by improper selection of entry points, and greatly improves the intelligence level, environmental adaptability and consistency of pesticide application quality of agricultural drone operations.

[0051] By traversing the spatial envelope of static obstacles, intersecting grid cells are accurately identified and their risk values ​​are set to the maximum, while being marked as impassable, effectively avoiding collision risks. Based on this, crop canopy height maps and canopy density analysis are introduced to identify the turbulent influence layer above the canopy within the height range, and a disturbance value is dynamically assigned based on the vegetation density of this area, forming a first risk value that accurately reflects airflow instability caused by vegetation disturbance. Further, combining the kernel density estimation results of pest and disease occurrence points, risk depression areas are identified, and a second risk value is generated by introducing a disease density guidance value, enabling path planning to have intelligent guidance capabilities that prioritize coverage of high-incidence areas. Simultaneously, a dynamic wake prediction zone is constructed based on real-time wind speed and direction, and wind-induced disturbance intensity is superimposed on passable grid cells falling into this zone to generate a third risk value, achieving proactive avoidance of vortex zones on the leeward side of obstacles. Finally, by integrating the spatial coordinates, resolution, and corresponding maximum or graded risk values ​​of all grid cells, a multi-dimensional risk field integrating geometric obstacles, airflow disturbances, crop status, and disease distribution is constructed. Based on this, the three-dimensional discrete grid field not only breaks through the limitations of traditional binary maps, but also significantly improves the level of intelligence in plant protection operations in complex farmland environments. Attached Figure Description

[0052] 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:

[0053] Figure 1A flowchart illustrating a method for monitoring pests and diseases using unmanned aerial vehicles (UAVs) provided in this application embodiment;

[0054] Figure 2 A structural diagram of a drone-based pest control system for monitoring pests and diseases provided in an embodiment of this application;

[0055] Figure 3 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0056] 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 intended to explain this application, and should not be construed as limiting this application.

[0057] The following describes an embodiment of the present application of a drone-based pest control method for monitoring pests and diseases, with reference to the accompanying drawings.

[0058] It should be noted that the execution subject of the drone control method for pest and disease monitoring in this application embodiment is a drone control system for pest and disease monitoring in this application embodiment. The drone control system for pest and disease monitoring can be configured in an electronic device so that the electronic device can perform a drone control function for pest and disease monitoring.

[0059] like Figure 1 As shown, this method for controlling pests and diseases using drones includes the following steps:

[0060] Step 1: Obtain the initial inspection path, control the UAV to fly along the initial inspection path, and collect crop canopy image sequences in real time.

[0061] In this embodiment, geographic information data of the target farmland is first acquired, including digital orthophoto maps (DOM), digital elevation models (DEM), and field boundary vector data. This data is then fused using a Geographic Information System (GIS) platform to identify flyable areas, obstacle distribution, and field boundaries. Based on this geographic information data, an equidistant parallel flight path planning algorithm is used to generate an initial inspection path covering the entire field. The flight direction is adjusted according to the prevailing wind direction to reduce crosswind influence. The flight altitude is set at 5 meters above the ground, and the flight speed is 4 m / s. The spacing between adjacent flight paths is determined to be 8 meters based on the field of view of the multispectral imaging equipment and the ground resolution requirements, ensuring complete coverage of the image sequence. Subsequently, a UAV equipped with a multispectral imaging device (such as MicaSense RedEdge-MX) is autonomously controlled to fly along this initial inspection path, acquiring real-time multispectral image sequences of the crop canopy during flight, including green, red, red-edge, and near-infrared bands.

[0062] Step 2: Determine the pest and disease occurrence points based on the canopy image sequence, and determine the pest and disease occurrence point set based on the pest and disease occurrence points.

[0063] In this embodiment, the crop canopy multispectral image sequence acquired by the UAV is first preprocessed, including radiometric correction, geometric correction, and stitching, to ensure the accuracy of subsequent analysis. Next, various vegetation indices (such as NDVI and NDRE) are calculated based on these images, and thresholds are set according to the ideal vegetation index ranges for specific crops at different growth stages. These thresholds are determined by analyzing the average vegetation index of healthy crops over the past few years, combined with the experience and knowledge of agricultural experts. For example, during the critical growth period of a crop, if the NDVI value is below 0.5 and the NDRE value is below 0.2, the area is considered to have potential pest and disease risks or malnutrition.

[0064] By comparing the vegetation index at each time point with a set threshold, abnormal areas significantly below the threshold are identified. For each detected abnormal area, its location coordinates and corresponding vegetation index features are extracted. Subsequently, a spatial clustering algorithm (such as DBSCAN) is used to analyze the location coordinates of all vegetation index abnormal areas, identifying highly dense clusters as pest and disease occurrence points. Finally, all identified pest and disease occurrence points are integrated to construct a pest and disease occurrence point set P={p1,p2,...,pn}, where each element pn represents the geographic coordinates of the nth pest and disease occurrence point, and n represents the total number of pest and disease occurrence points.

[0065] Step 3: Based on the set of pest and disease occurrence points, determine several risk areas; determine the target hovering entry point based on each risk area.

[0066] First, kernel density estimation was performed on the spatial distribution of all disease points to quantify the degree of disease point clustering per unit area. A Gaussian kernel function was used, with the bandwidth parameter set to 1.5 meters based on field scale and disease propagation characteristics, to calculate a continuous spatial density function, which was then used to obtain the disease point density per square meter at each geographic coordinate. Subsequently, a density threshold of 0.8 points / m was set. 2 This threshold, determined based on historical disease outbreak data and field validation experiments, is used to distinguish between high-risk and low-risk areas. Connected regions of the spatial density function are identified as high-risk sub-regions and labeled Ri, representing the i-th risk region. Each risk region corresponds to an independent high-incidence patch. Morphological closing operations and region growing algorithms are used to smooth and optimize the initially divided regions, eliminating sporadic noise and ensuring clear and continuous boundaries for each risk region. Finally, several spatially separated risk sub-regions are generated.

[0067] Step 31: Determine the density peak point set of pest and disease occurrence points based on the risk area.

[0068] This embodiment determines the observed peak point set based on the density of disease spots calculated using a spatial density function. Local maximum points, i.e., density peak points, are searched within each risk area. For complex lesion areas with multiple density peaks, all valid peak points are retained to reflect the potential multi-center diffusion characteristics of the disease.

[0069] Step 32: Determine the geometric center based on the risk area, and construct a guide ray based on the nearest dense peak point of the geometric center.

[0070] The geometric center is the arithmetic mean of all geographic coordinates within the current risk area boundary. The Euclidean distance from the geometric center to each density peak point is calculated, and the density peak point with the shortest distance is selected as the nearest core point (the nearest dense peak point). A guiding ray is constructed from the geometric center towards the nearest core point.

[0071] Step 33: Based on the guiding ray, construct multiple candidate anchor points.

[0072] On the guide ray, multiple candidate anchor points for luminaires are set at intervals d. The interval d ranges from 0.5 to 2m, and is preferably 1m in this embodiment.

[0073] Step 34: Preset the spraying operation height; determine the crop canopy height map based on the crop canopy image sequence, and determine the average height of the crop canopy below each candidate anchor based on the crop canopy height map; determine the height difference based on the spraying operation height and the average height of the crop canopy, and determine whether the height difference is greater than or equal to 0.3m and less than or equal to 0.5m. If yes, proceed to the next step; if no, end.

[0074] In this embodiment, the preset drone spraying height is 0.4 meters above the crop canopy, meaning the vertical distance between the target flight height and the top of the crop below should be maintained within the range of 0.3 to 0.5 meters to ensure uniform droplet settling and reduce drift. Based on the multispectral image sequence acquired in step 1 and the simultaneously acquired LiDAR point cloud data, a high-resolution crop canopy height map is constructed to represent the absolute elevation of the top of the vegetation canopy or its height relative to the ground at a horizontal position (x, y). For each candidate anchor point generated in step 33, the canopy height value of the corresponding area directly below it is extracted, and the average height of all positions within that area is calculated as the average height of the crop canopy below that anchor point. Subsequently, the height difference between the spraying operation height and the actual canopy height is calculated. It is determined whether the height difference is greater than or equal to 0.3m and less than or equal to 0.5m. If it meets the condition, the candidate anchor point is retained and proceeds to the next safety verification step; otherwise, the flight height at that position is considered unsuitable, which may lead to pesticide damage due to spraying too close or droplet drift due to spraying too high, so the candidate point is removed. This screening mechanism ensures that the final selected hovering entry point can achieve both precise pesticide application and compliance with agronomic operation standards.

[0075] Step 35: Obtain a 3D point cloud map of the work area, and extract the spatial envelope of static obstacles based on the 3D point cloud map.

[0076] In this embodiment, the LiDAR sensor onboard the UAV is first used to perform a comprehensive scan of the work area, obtaining a three-dimensional point cloud map composed of a large number of discrete spatial coordinate points. Each point represents the position information of the surface of the object being measured. Through the dense distribution of points, the terrain and the three-dimensional shapes of various objects within the work area can be accurately depicted.

[0077] The raw point cloud data is preprocessed using filtering algorithms (such as voxel mesh filtering and statistical outlier removal) to eliminate noise and unnecessary details while preserving key structural features. Then, clustering analysis techniques (such as Euclidean distance clustering or region growing algorithms) are employed to group points belonging to the same physical entity together, thereby identifying potential static obstacles. For each detected obstacle, its minimum bounding geometry (such as a cube, cylinder, or polygonal prism) is further calculated as its spatial envelope model. This process involves determining the obstacle's boundary point set and using mathematical methods to solve for the geometry that completely encloses this point set while minimizing its volume. Ultimately, the generated static obstacle spatial envelope not only contains the approximate location and size information of the obstacles but also provides their precise layout in three-dimensional space.

[0078] Step 36: Based on the candidate anchor point, project it vertically onto the ground to obtain a horizontal projection point; construct a search space with the horizontal projection point as the center. If the search space does not have a static obstacle spatial envelope, then end; if the search space has one or more static obstacle spatial envelopes, then proceed to the next step.

[0079] A vertical column search space is constructed centered on the horizontal projection point. This search space is a circular area on the horizontal plane, with a radius preferably set to 2.0m; in the vertical direction, the column height extends from the ground to the sum of the preset spraying operation height and the safety margin (0.5m). Therefore, this column space simulates the three-dimensional envelope area that the UAV may occupy when hovering at the target point.

[0080] By comparing the previously acquired static obstacle spatial envelope data, we check whether any static obstacles exist within the search space. If there is no spatial envelope containing any static obstacles within the search space, it indicates that the location is safe and can be directly used for the next step of path planning or task deployment. At this point, the processing of this candidate anchor point ends.

[0081] However, if the search space contains the spatial envelope of one or more static obstacles, it means that there is a potential risk around the currently selected horizontal projection point, and the safe operation of the UAV in this area cannot be guaranteed. In this case, further measures are required, and this embodiment selects to perform step 37.

[0082] Step 37: Real-time acquisition of wind speed and direction; determination of the prevailing wind direction vector based on the wind direction; determination of the first obstacle based on static obstacles; formation of a wake prediction zone by extending along the opposite direction of the prevailing wind direction vector to the leeward side based on the first obstacle; termination if the horizontal projection point is located in the wake prediction zone; termination if the horizontal projection point is not located in the wake prediction zone, using the current candidate anchor point as the target hovering entry point.

[0083] In this embodiment, wind speed and direction information of the work area are acquired in real time. The current wind speed (3.2 m / s) and wind direction (135°) are collected using a miniature meteorological sensor mounted on a drone or a ground-based wireless anemometer. The prevailing wind direction vector is determined based on the wind direction data. Subsequently, combined with the static obstacle spatial envelope extracted in step 35, obstacles with a height greater than 2.0 m are selected as the first obstacle. For example, irrigation equipment supports or field fence posts with a height of 2.5 m are sufficient to generate a significant wake disturbance zone downwind.

[0084] For each identified first obstacle, a wake prediction zone is constructed by extending it in the opposite direction of the prevailing wind vector, starting from its leeward side (i.e., the side opposite to the wind direction). This zone uses a conical spatial model, starting at the center of the obstacle's leeward side, with an angle of 60°. Its length is set according to the empirical formula for wind speed as L=5×h (where h is the obstacle height). For example, for an obstacle 2.5m high, the wake zone extends to approximately 12.5m, forming a high-risk area for airflow disturbance in the downwind direction.

[0085] Next, it is determined whether the horizontal projection point corresponding to the candidate anchor point generated in step 36 falls within any wake prediction region. If the horizontal projection point of a candidate anchor point is located within the conical region, it is considered to be in an unstable airflow region. The UAV is susceptible to vortices and gusts when hovering here, posing a risk of loss of flight attitude control. Therefore, the verification process for this candidate point is terminated. If it does not fall within any wake prediction region, it indicates that the airflow environment at this point is relatively stable, and the current candidate anchor point can be used as the target hovering entry point.

[0086] If multiple candidate anchor points are evaluated, and if multiple horizontal projection points are located in the wake prediction region, the first candidate anchor point that passes the verification will be used as the target hovering entry point.

[0087] In this embodiment, dense peak points of pests are identified through kernel density estimation, and candidate anchor points are generated by constructing a guide ray based on the geometric center of the risk area. This breaks through the simplistic logic of "center is target" or "nearest point is optimal" in traditional path planning. This guidance mechanism, which integrates spatial distribution characteristics, allows the drone hovering point to accurately focus on the core hotspots of pest outbreaks, thereby improving the targeting and pesticide utilization efficiency of control and avoiding the waste of resources from large-scale coverage and inefficient spraying. At the same time, the design of the guide ray takes into account the overall shape of the area, preventing it from falling into the margins or isolated points due to excessive pursuit of local peaks.

[0088] Furthermore, the introduction of multi-level screening mechanisms, including crop canopy height matching, three-dimensional obstacle avoidance, and wind wake prediction, marks a leap from "being able to fly" to "intelligent flight" in drone operations. Canopy height matching ensures that the spraying height is always maintained above the crops to avoid collisions; the optimal range of 0.3-0.5m in the three-dimensional obstacle space avoids both pesticide drift and insufficient deposition, and prevents the drone from being affected by terrain undulations. Envelope detection effectively avoids static obstacles such as power poles and irrigation facilities, improving flight safety; and the dynamic judgment of the wake prediction zone based on real-time wind field construction is the first time that micro-meteorological factors have been incorporated into the hovering point decision system, preventing the drone from getting caught in turbulent vortex zones on the leeward side of obstacles, leading to attitude instability or even crashes.

[0089] Step 4: Obtain the current position and environmental information of the UAV. Based on the target hovering entry point, the current position and environmental information of the UAV, construct a three-dimensional discrete grid field. The three-dimensional discrete grid field includes several grid cells and corresponding grid risk values.

[0090] A three-dimensional discrete grid field is constructed based on the target hovering entry point, the current position of the UAV, and environmental information, including:

[0091] Step 41: Divide the work area into a three-dimensional discrete grid; each grid cell of the three-dimensional discrete grid includes a first coordinate and a resolution.

[0092] In this embodiment, the entire farmland space is first divided into a three-dimensional discrete raster map based on the geographical boundaries and terrain elevation data of the work area. The horizontal extent of this three-dimensional space covers all identified high-risk work sub-areas and their surrounding buffer zones, while the vertical direction extends from the ground (or the lowest canopy height) to the maximum safe flight altitude of the UAV (e.g., 5 meters above the ground). Each raster cell is a cube with uniform side length and a resolution of 0.5m to achieve a balance between computational efficiency and spatial accuracy. Each raster cell has unique three-dimensional coordinates (X, Y, Z), where X and Y represent the horizontal position and Z represents the vertical layer height, used to identify its specific location in space. All rasters constitute a three-dimensional array structure for subsequent storage and retrieval of attribute information for each spatial location.

[0093] Step 42: Initialize the grid risk value to 0 for each grid cell.

[0094] Step 43: Traverse the spatial envelope of all static obstacles. If an obstacle intersects with the current grid cell, mark the grid risk value of the current grid cell as the maximum risk value and mark the current grid cell as impassable.

[0095] In this embodiment, a three-dimensional discrete grid is traversed based on the spatial envelope of static obstacles. For each static obstacle, it is determined whether its spatial envelope geometrically intersects with the current grid cell: if the cubic space of the grid cell overlaps with the envelope of any obstacle, the grid is considered to be located inside the obstacle or occupied by it. At this time, the grid risk value of the grid cell is assigned to a preset maximum risk value (preferably 100), and the grid is marked as impassable.

[0096] Step 44: Based on the grid cells with unmarked impassable states and the vertical coordinate Z of the current grid cell satisfying: hc(X,Y)+0.3≤Z≤hc(X,Y)+0.8, it is determined that the current grid cell is located in the turbulent influence layer above the canopy; if not satisfied, the grid risk value remains unchanged; for the turbulent influence layer above the canopy, the canopy density of the grid cells with unmarked impassable states is determined based on the crop canopy height map, the perturbation value is determined based on the canopy density, and the first risk value is obtained based on the grid risk value and the perturbation value; where hc(X,Y) is the absolute elevation of the crop canopy height map at the horizontal position (X,Y).

[0097] The range hc(X,Y)+0.3≤Z≤hc(X,Y)+0.8 is defined as the turbulence influence layer above the canopy, specifically the area 0.3 to 0.8 meters above the top of the crop canopy. Within this region, irregular airflow disturbances are easily generated due to vegetation surface friction and wind shearing, affecting the flight stability of the UAV. If the Z-coordinate of a grid cell falls within this range, the turbulence risk assessment process begins; otherwise, its grid risk value remains unchanged.

[0098] Canopy density is obtained through interpolation of vegetation point cloud density or Normalized Difference Vegetation Index (NDVI) per unit area, reflecting the degree of vegetation density. Canopy density is mapped to a perturbation value between 10 and 40: for example, a perturbation value of 10 corresponds to sparse canopy (density <30%), 25 corresponds to medium density (30%-50%), and 40 corresponds to high density (>50%). This perturbation value is then superimposed on the current raster risk value to obtain the first risk value.

[0099] Step 45: If, based on the grid cells that are not marked as impassable, the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is greater than the target threshold, the current grid cell is determined to be located in a risk depression; if the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is less than or equal to the target threshold, the grid risk value remains unchanged; for risk depressions, a guiding value is determined based on the density of pest and disease occurrence points, and a second risk value is obtained based on the guiding value and the grid risk value.

[0100] When the density of pest and disease occurrence points in a certain grid cell located in a risk depression (X,Y) exceeds the target threshold, a guiding value is further determined based on the density of pest and disease occurrence points within that cell. This guiding value, ranging from 20 to 30, reflects the guiding impact of the severity of the disease in that area on path planning. To achieve quantitative assessment, a mapping relationship is established between the density of the disease and the guiding value: when the density is between 90 and 95, the guiding value increases linearly with the density; specifically, when the density is 90, the guiding value is the minimum of 20; when the density reaches 95-100, the guiding value is the maximum of 30. This mapping mechanism ensures that the guiding value increases smoothly with the severity of the disease, highlighting the priority of high-density areas while avoiding excessive amplification of risk bias. Subsequently, the calculated guiding value is added to the original risk value of the current grid cell to obtain a second risk value. If the second risk value is less than 0, it is marked as 0.

[0101] Step 46: For grid cells located in the wake prediction zone that are not marked as impassable, determine the wind-induced disturbance volume based on the real-time wind speed, and obtain the third risk value based on the wind-induced disturbance volume and the grid risk value.

[0102] Based on real-time acquired wind speed data, the wind-induced disturbance intensity is calculated. This intensity reflects the degree of airflow instability caused by vortices and turbulence on the leeward side of the obstacle. The wind-induced disturbance intensity is converted into wind-induced disturbance airflow rate through an empirical mapping function, with a value range of [10, 50]. The mapping relationship is as follows: when the wind speed is less than 2 m / s, the wind-induced disturbance airflow rate is 10; when the wind speed is above 50, the wind-induced disturbance airflow rate is 10.

[0103] For wind speeds between 2 and 4 m / s, linear interpolation is used to calculate the disturbance value. When the wind speed is greater than 4 m / s, the wind-induced disturbance volume is set to the maximum value of 50, indicating entry into the strong turbulence hazard zone. Subsequently, the calculated wind-induced disturbance volume is superimposed on the existing grid risk value of the current grid cell to obtain the third risk value.

[0104] Step 47: Based on all grid cells and their corresponding first coordinates, resolution, grid risk value or maximum risk value or first risk value or second risk value or third risk value, a three-dimensional discrete grid field is constructed.

[0105] The grid field is based on the three-dimensional spatial grid divided in step 41. Each grid cell has unique three-dimensional coordinates (X,Y,Z) and a preset resolution, and carries the risk value determined in step 46. Specifically, the attributes of each grid cell include: horizontal and vertical position information, whether it is marked as impassable, and the final determined grid risk value. This risk value is dynamically assigned according to its environment: if it is located inside a static obstacle, it has the maximum risk value (100); if it is in the canopy turbulence influence layer, it has a disturbance value superimposed to form a first risk value; if it is located in a high-risk depression with pests and diseases, it has a guiding value further added to form a second risk value; if it is in the wake prediction zone and the wind speed is high, it has a wind-induced disturbance wind volume superimposed to form a third risk value. All grid cells are organized into a three-dimensional array or sparse voxel map structure according to spatial index, forming a three-dimensional risk grid field with full coverage and complete attributes.

[0106] Step 5: Within the three-dimensional discrete grid field, multiple candidate trajectories are generated based on the RRT algorithm, and the transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis.

[0107] In this embodiment, based on the constructed 3D discrete grid field, a Fast Exploratory Random Tree (RRT) algorithm is used to generate multiple candidate flight trajectories. The starting position of the UAV is taken as the root node of the tree, and the target area (target hovering entry point) is set as the guiding direction. Reachable points are randomly sampled in the 3D grid space, and a guidance strategy is used to improve the efficiency of expanding to high-priority areas. During each sampling, the nearest node is selected from the leaf node set of the current tree, and the path is expanded one step towards the random point or target point with a step size of 1.0 meter, generating a new intermediate node. During the expansion process, continuous collision detection is performed on all grid cells traversed by the newly generated path segment: if any grid is marked as impassable or its comprehensive risk value exceeds a preset safety threshold, the path segment is determined to be infeasible, and the expansion is abandoned; otherwise, the new node is added to the search tree. To improve path diversity, the algorithm runs in parallel multiple times, each time using a different random seed and target guidance weight, ultimately generating a set of candidate trajectories with varying spatial distributions. Each trajectory consists of a series of ordered 3D grid coordinate sequences, representing a feasible flight path from the starting point to the target area.

[0108] The transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis, including:

[0109] Step 51: Based on each candidate trajectory, extract the sequence of all raster cells that have passed through it.

[0110] Step 52: Determine the target risk value based on the grid cell sequence.

[0111] The target risk value is the maximum risk value in the raster cell sequence.

[0112] Step 53: If the target risk value is greater than or equal to the preset safety threshold, the current candidate trajectory is eliminated; if the target risk value is less than the preset safety threshold, the current candidate trajectory is retained and marked as the expected trajectory.

[0113] The safety threshold value is 90.

[0114] Step 54: Preset target wind speed range and target wind direction range, wherein the target wind speed range includes several wind speeds and the target wind direction range includes several wind directions; based on the several wind speeds in the target wind speed range and the several wind directions in the target wind direction range, construct multiple wind field scenario combinations.

[0115] In this embodiment, the preset target wind speed range is [2.0 m / s, 4.5 m / s], encompassing several typical wind speed values ​​(e.g., 2.0, 2.5, 3.0, 3.5, 4.0, 4.5 m / s), and the target wind direction range is [10°, 180°], covering multiple key directions from northeast to southwest winds (e.g., 10°, 45°, 90°, 135°, 180°). Based on the discrete values ​​of these two ranges, multiple wind field scenario combinations are constructed. Each combination consists of a specific pairing of wind speed and wind direction, such as (v=4.0 m / s, θ=135°), resulting in a total of 6 × 5 = 30 basic wind field scenarios. This embodiment also introduces a wind speed change rate of [−0.5, 0.5] m / s. 2

[0116] The wind direction change rate [−15° / s, 15° / s] is used to simulate the sudden and gradual changes in the wind field within a preset time interval [5, 15] s, generating 10 sets of wind field scenarios, each corresponding to a different combination of wind field scenarios.

[0117] Step 55: Based on each wind field scenario combination and the real-time wind speed and direction, obtain the first wind speed and first wind direction for the preset time interval.

[0118] Step 56: Using the first wind speed and the first wind direction, generate the first wake prediction zone for each wind field scenario combination and readjust the grid risk value of the three-dimensional discrete grid field.

[0119] Although a wake prediction region was generated and risk assigned during the construction of the 3D discrete grid field, this initial state only reflects the instantaneous meteorological conditions at the current moment. Due to the significant time-varying and uncertain nature of wind fields, relying solely on static modeling will result in path planning lacking the ability to proactively adapt to future airflow disturbances. Therefore, it is necessary to dynamically reconstruct the wake prediction region for each wind field scenario combination based on the first wind speed and first wind direction within a preset time interval, and to reassess its impact on flight space.

[0120] Specifically, for each wind field scenario, the spatial extent of the wake prediction area is recalculated using the corresponding first wind speed and first wind direction:

[0121] L wake (t) represents the length of the wake prediction region at time t. Represents the first empirical coefficient. The value is 1.2, where vfirst represents the first wind speed. Represents the duration of the eddy current. It takes 3 seconds.

[0122] , Wwake(t) represents the width of the wake prediction region at time t. Represents the second empirical coefficient. The value is 1.5, where Wobs represents the projected width of the obstacle in the vertical wind direction. This represents the angle between the first wind direction and the principal axis direction of the static obstacle. Representing the first wind direction, Represents the main axis direction of a static obstacle. Represents the included angle coefficient. It is 0.03.

[0123] Subsequently, with the leeward side of the static obstacle as the center, a wake prediction area in the form of a cone or rectangular column is reconstructed in the opposite direction of the updated prevailing wind. For all grid cells that fall into this area and are not marked as impassable, the wind-induced disturbance intensity is mapped according to the first wind speed, and this value is superimposed on the original grid risk value to form a new third risk value (i.e., the grid risk value of the three-dimensional discrete grid field is readjusted).

[0124] Step 57: Based on the readjusted grid risk value, determine the risk exposure of the UAV flying under each wind field scenario combination, and based on the risk exposure, select transition trajectories from the expected trajectories.

[0125] The expression for risk exposure is as follows:

[0126] E represents risk exposure. The estimated flight time required for a UAV to execute a specific candidate trajectory is preferably 60-300 seconds, C. eff (t) represents the risk value of the grid cell where the UAV is located at time t, and wrisk represents the sensitivity coefficient, which is preferably 1.2.

[0127] Step 571: Determine whether the risk exposure is greater than the target exposure value.

[0128] Determining the risk exposure level comprehensively considers the drone platform's wind resistance, mission priority, battery redundancy, and the maximum acceptable risk level from historical flight data. For example, for high-stability models, the target exposure value can be set higher; while for small multi-rotor drones or complex canopy environments, the target exposure value should be more conservative.

[0129] Step 572: Locate the last grid cell whose expected trajectory is less than or equal to the safety threshold and mark it as the midpoint.

[0130] The midpoint represents the farthest reachable position within the current expected trajectory that is still in a safe flight state, and serves as the starting reference point for subsequent trajectory adjustments.

[0131] Step 573: Determine the distance between the midpoint and the expected trajectory endpoint, and mark it as the adjustment distance.

[0132] After determining the intermediate point, calculate the Euclidean distance between that point and the original expected trajectory endpoint, denoted as the adjustment distance. This reflects the remaining flight mission that still needs to be completed after the current trajectory is interrupted.

[0133] Step 574: Determine whether the adjustment distance is less than or equal to the minimum correction distance.

[0134] The minimum correction distance is preferably 3 meters. When the adjustment distance is less than or equal to 3 meters, it indicates that the planned trajectory of the UAV has a short remaining flight path after entering the high-risk area, and it is already close to the target hovering point. In this case, if the UAV is required to complete heading adjustment, speed control, and attitude stabilization within such a short distance, and reposition itself to a new safe hovering position, it will be difficult to achieve the desired result.

[0135] Step 575: If the judgment result of step 571 and / or step 574 is yes, then the current expected trajectory is removed; otherwise, the expected trajectory is defined as a transition trajectory.

[0136] If step 571 is true (risk exposure is too high) or step 574 is true (adjustment distance is insufficient), the expected trajectory is determined to be unsafe to execute and is removed from the candidate set; if neither condition is met (i.e. the risk is controllable and there is sufficient room for correction), the trajectory is defined as a transition trajectory.

[0137] When multiple transition trajectories exist, executing any one of them can safely reach the target hovering entry point and complete the hovering operation. All transition trajectories meet dual safety constraints: they pass the global safety verification of risk exposure and possess sufficient path correction redundancy, meaning each transition trajectory is within an acceptable risk window. Regardless of which trajectory is chosen, the environmental risk accumulated by the UAV throughout the flight is lower than the target exposure value, and it has sufficient adjustment distance for attitude adjustment and precise positioning when approaching the target area. Furthermore, although these trajectories differ geometrically, their endpoints all converge to the same target hovering entry point, and the paths avoid impassable areas and high-risk wake core areas throughout.

[0138] After executing the selected transition trajectory, the drone smoothly flies along the path to the target hovering entry point, entering a stable hovering state at a preset altitude above the crop canopy. At this point, the onboard spraying system automatically starts according to preset operating parameters, dynamically adjusting nozzle flow rate, droplet size, and spray width based on real-time wind speed, direction, and crop density information to ensure uniform coverage of areas with high pest and disease incidence. Simultaneously, the flight control system continuously monitors airflow disturbances and positional deviations, fine-tuning attitude through closed-loop control to maintain hovering accuracy. After completing the set duration of precise spraying, the drone autonomously exits the hovering state according to the planned path to the next target point and enters the next operating segment.

[0139] like Figure 2 As shown, this embodiment also discloses a drone-based pest control system for monitoring pests and diseases, comprising the following modules:

[0140] Inspection module: used to obtain the initial inspection path, control the UAV to fly along the initial inspection path, and collect crop canopy image sequences in real time;

[0141] Pest and disease occurrence point determination module: used to determine the occurrence points of pests and diseases based on the canopy image sequence, and to determine the set of occurrence points of pests and diseases based on the occurrence points of pests and diseases;

[0142] Target hovering entry point construction module: used to determine several risk areas based on the set of pest and disease occurrence points; and to determine a target hovering entry point based on each risk area.

[0143] 3D Discrete Grid Field Construction Module: Used to acquire the current position and environmental information of the UAV, and construct a 3D discrete grid field based on the target hovering entry point, the current position and environmental information of the UAV. The 3D discrete grid field includes several grid cells and corresponding grid risk values.

[0144] Transition trajectory construction module: used to generate multiple candidate trajectories based on the RRT algorithm within the three-dimensional discrete grid field, and determine the transition trajectory based on the target hovering entry point and grid risk value, combined with wind field analysis.

[0145] To implement the above embodiments, this application also proposes an electronic device. Please see [link to relevant documentation]. Figure 3 , Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of this application. For example... Figure 3 As shown, the electronic device 500 includes: a processor 501 and a memory 502 communicatively connected to the processor 501; the memory 502 stores computer-executable instructions; the processor 501 executes the computer-executable instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0146] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0147] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0148] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A drone-based pest control method for monitoring pests and diseases, characterized in that, Includes the following steps: Step 1: Obtain the initial inspection path, control the drone to fly along the initial inspection path, and collect crop canopy image sequences in real time; Step 2: Determine the pest and disease occurrence points based on the canopy image sequence, and determine the pest and disease occurrence point set based on the pest and disease occurrence points; Step 3: Based on the set of pest and disease occurrence points, determine several risk areas; Determine the target hovering entry point based on each of the aforementioned risk areas; Step 4: Obtain the current position and environmental information of the UAV. Based on the target hovering entry point, the current position and environmental information of the UAV, construct a three-dimensional discrete grid field. The three-dimensional discrete grid field includes several grid cells and corresponding grid risk values. Step 5: Within the three-dimensional discrete grid field, multiple candidate trajectories are generated based on the RRT algorithm, and the transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis.

2. The method for controlling pests and diseases using unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Determining the target hovering entry point based on each of the aforementioned risk areas includes the following steps: Step 31: Based on the set of pest and disease occurrence points in the risk area, kernel density estimation is performed to obtain the density of pest and disease occurrence points, and the set of density peak points is determined based on the density of pest and disease occurrence points. Step 32: Determine the geometric center based on the risk area, and construct a guide ray based on the nearest dense peak point of the geometric center; Step 33: Based on the guiding ray, construct multiple candidate anchor points; Step 34: Preset the spraying operation height; determine the crop canopy height map based on the crop canopy image sequence, and determine the average crop canopy height below each candidate anchor based on the crop canopy height map; determine the height difference based on the spraying operation height and the average crop canopy height, and determine whether the height difference is greater than or equal to 0.3m and less than or equal to 0.5m. If yes, proceed to the next step; if no, end. Step 35: Obtain a 3D point cloud map of the work area, and extract the spatial envelope of static obstacles based on the 3D point cloud map; Step 36: Based on the candidate anchor points, project them vertically onto the ground to obtain horizontal projection points; construct a search space with the horizontal projection points as the center. If the search space does not contain a static obstacle space envelope, the process ends; if the search space contains one or more static obstacle space envelopes, proceed to the next step. Step 37: Real-time acquisition of wind speed and direction; determination of the prevailing wind direction vector based on the wind direction; determination of the first obstacle based on static obstacles; formation of a wake prediction zone by extending along the opposite direction of the prevailing wind direction vector to the leeward side based on the first obstacle; termination if the horizontal projection point is located in the wake prediction zone; termination if the horizontal projection point is not located in the wake prediction zone, using the current candidate anchor point as the target hovering entry point.

3. The method for controlling pests and diseases using unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, If multiple horizontal projection points are located within the wake prediction region, the first verified candidate anchor point will be used as the target hovering entry point.

4. A method for pest and disease monitoring using unmanned aerial vehicles (UAVs), as described in claim 2, is characterized in that... A three-dimensional discrete grid field is constructed based on the target hovering entry point, the current position of the UAV, and environmental information, including: Step 41: Divide the work area into a three-dimensional discrete grid; each grid cell of the three-dimensional discrete grid includes a first coordinate and a resolution; Step 42: Initialize the grid risk value to 0 for each grid cell; Step 43: Traverse the spatial envelope of all static obstacles. If an obstacle intersects with the current grid cell, mark the grid risk value of the current grid cell as the maximum risk value and mark the current grid cell as impassable. Step 44: Based on the grid cells without marked impassable states and the vertical coordinate Z of the current grid cell satisfying: hc(X,Y)+0.3≤Z≤hc(X,Y)+0.8, it is determined that the current grid cell is located in the turbulent influence layer above the canopy; if not, the grid risk value remains unchanged; for the turbulent influence layer above the canopy, the canopy density of the grid cells without marked impassable states is determined based on the crop canopy height map, the perturbation value is determined based on the canopy density, and the first risk value is obtained based on the grid risk value and the perturbation value; where hc(X,Y) is the absolute elevation of the crop canopy height map at the horizontal position (X,Y); Step 45: If, based on grid cells that are not marked as impassable, the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is greater than the target threshold, the current grid cell is determined to be located in a risk depression; if the density of pest and disease occurrence points at the horizontal position (X,Y) of the current grid cell is less than or equal to the target threshold, the grid risk value remains unchanged; for risk depressions, a guiding value is determined based on the density of pest and disease occurrence points, and a second risk value is obtained based on the guiding value and the grid risk value; Step 46: For grid cells located in the wake prediction area that are not marked as impassable, determine the wind-induced disturbance volume based on the real-time wind speed, and obtain the third risk value based on the wind-induced disturbance volume and the grid risk value. Step 47: Based on all grid cells and their corresponding first coordinates, resolution, grid risk value or maximum risk value or first risk value or second risk value or third risk value, a three-dimensional discrete grid field is constructed.

5. A method for monitoring pests and diseases using unmanned aerial vehicles (UAVs), as described in claim 4, is characterized in that... If the second risk value is less than 0, then the second risk value is marked as 0.

6. A method for pest and disease monitoring using unmanned aerial vehicles (UAVs), as described in claim 2, is characterized in that... The transition trajectory is determined based on the target hovering entry point and grid risk value, combined with wind field analysis, including: Step 51: Based on each candidate trajectory, extract the sequence of all raster cells passed through; Step 52: Determine the target risk value based on the grid cell sequence; Step 53: If the target risk value is greater than or equal to the preset safety threshold, the current candidate trajectory is eliminated; if the target risk value is less than the preset safety threshold, the current candidate trajectory is retained and marked as the expected trajectory. Step 54: Preset target wind speed range and target wind direction range, wherein the target wind speed range includes several wind speeds and the target wind direction range includes several wind directions; based on the several wind speeds in the target wind speed range and the several wind directions in the target wind direction range, construct multiple wind field scenario combinations. Step 55: Based on each wind field scenario combination and the real-time wind speed and direction, obtain the first wind speed and first wind direction for a preset time interval; Step 56: Using the first wind speed and the first wind direction, generate the first wake prediction area for each wind field scenario combination and readjust the grid risk value of the three-dimensional discrete grid field. Step 57: Based on the readjusted grid risk value, determine the risk exposure of the UAV flying under each wind field scenario combination, and based on the risk exposure, select transition trajectories from the expected trajectories.

7. A method for pest and disease monitoring using unmanned aerial vehicles (UAVs), as described in claim 6, is characterized in that... Based on risk exposure, transition trajectories are selected from the expected trajectories, including: Step 571: Determine whether the risk exposure is greater than the target exposure value; Step 572: Locate the last grid cell whose expected trajectory is less than or equal to the safety threshold and mark it as the midpoint; Step 573: Determine the distance between the midpoint and the expected trajectory endpoint, and mark it as the adjustment distance; Step 574: Determine whether the adjustment distance is less than or equal to the minimum correction distance; Step 575: If the judgment result of step 571 and / or step 574 is yes, then the current expected trajectory is removed; otherwise, the expected trajectory is defined as a transition trajectory.

8. A drone-based pest control system for monitoring pests and diseases, used to execute the drone-based pest control method for monitoring pests and diseases as described in any one of claims 1-7, characterized in that, Includes the following modules: Inspection module: used to obtain the initial inspection path, control the UAV to fly along the initial inspection path, and collect crop canopy image sequences in real time; Pest and disease occurrence point determination module: used to determine the occurrence points of pests and diseases based on the canopy image sequence, and to determine the set of occurrence points of pests and diseases based on the occurrence points of pests and diseases; Target hovering entry point construction module: used to determine several risk areas based on the set of pest and disease occurrence points; Determine the target hovering entry point based on each of the aforementioned risk areas; 3D Discrete Grid Field Construction Module: Used to acquire the current position and environmental information of the UAV, and construct a 3D discrete grid field based on the target hovering entry point, the current position and environmental information of the UAV. The 3D discrete grid field includes several grid cells and corresponding grid risk values. Transition trajectory construction module: used to generate multiple candidate trajectories based on the RRT algorithm within the three-dimensional discrete grid field, and determine the transition trajectory based on the target hovering entry point and grid risk value, combined with wind field analysis.

Images