Targeted molluscicidal method and system based on unmanned aerial vehicle group cooperation and AI recognition

Abstract

The invention discloses a precise targeted snail control method and system based on unmanned aerial vehicle cluster cooperation and AI recognition, and belongs to the technical field of unmanned aerial vehicle cluster control environment governance. The method comprises the following steps: receiving and analyzing operation area information through a cloud control center, automatically generating a global collaborative investigation path, and distributing the global collaborative investigation path to an investigation unmanned aerial vehicle group; the unmanned aerial vehicle group synchronously collects ground multi-source image data according to a path, and identifies oncomelania breeding land targets and geographic coordinates thereof in real time through an airborne AI model; the cloud end collects all recognition results, an improved DBSCAN algorithm is adopted for space-time clustering and data fusion, a snail control target point electronic map is generated, and pesticide application tasks are dynamically distributed; the pesticide applying unmanned aerial vehicle performs centimeter-level accurate positioning through RTK positioning, and controls the spraying mechanism to perform instantaneous quantitative pesticide spraying on the target center; and finally, the snail extinguishing effect is automatically verified through secondary scanning and AI image comparison. According to the method, full-process automation of investigation, identification, decision making, spraying and verification is achieved, the snail killing efficiency and accuracy are greatly improved, the dosage of drugs is remarkably reduced, and the risk of personnel infection is completely eradicated.

Description

Technical Field

[0001] This invention relates to the field of drone swarm control environment management technology, and in particular to a targeted snail eradication method and system based on drone swarm collaboration and AI recognition. Background Technology

[0002] Oncomelania snails are the only intermediate host for schistosomiasis, and their effective control is crucial to blocking the spread of the disease. Currently, the main method relies on manual snail detection and pesticide spraying, which has the following inherent drawbacks:

[0003] Inefficient: Manual surveying is slow and arduous, making it difficult to cope with complex terrains such as swamps, water-filled rice paddies, and has a limited coverage area.

[0004] Inaccurate identification: It is very easy to miss snails by visual inspection, especially for juvenile snails or snails hidden under vegetation, which leads to inaccurate determination of breeding grounds.

[0005] Severe pollution: To ensure effectiveness, the entire area is usually sprayed, resulting in more than 95% of the agent not reaching the target, causing serious environmental drug pollution.

[0006] High risk: Staff are exposed to contaminated water and chemical agents for extended periods, posing a significant health risk.

[0007] While existing technologies have seen attempts at single-drone applications, they mostly employ a separate model of "aerial photography followed by manual interpretation and then operation," failing to create a real-time closed loop and exhibiting low levels of intelligence. This makes them unsuitable for large-scale, high-efficiency snail eradication. Therefore, a completely new integrated and intelligent solution is urgently needed. Summary of the Invention

[0008] The purpose of this invention is to propose a targeted snail eradication method based on drone swarm collaboration and AI recognition, aiming to solve the problems of low efficiency, inaccurate identification, serious waste of drugs, and high personal safety risks of existing snail eradication methods that rely on manual labor.

[0009] The present invention is implemented as follows: a targeted snail eradication method based on drone swarm collaboration and AI recognition includes the following steps: Receive and parse the work area information and initialize the parameters, generate a global collaborative exploration path, and distribute it to each exploration drone; The drone swarm is instructed to fly along the global collaborative reconnaissance path, simultaneously collect multi-source raw ground image data, and perform real-time analysis based on a pre-trained AI model to identify snail breeding ground targets and their geographical coordinates. It receives all identified target information data, performs data fusion and deduplication; generates a global electronic map of mollusc eradication targets; and dynamically allocates application tasks to the optimal application drone. The assigned spraying drones are precisely positioned and the spraying mechanism is controlled to spray the target area instantaneously and quantitatively. After the spraying is completed, a reconnaissance drone is assigned to perform a second scan of the sprayed area. By comparing the changes in the images before and after the spraying with AI, the snail eradication effect is automatically verified and an operation report is generated.

[0010] Another objective of this invention is to propose a targeted snail eradication method system based on drone swarm collaboration and AI recognition. The system includes:

[0011] The task planning and cluster initialization module is used to receive work area information, parse and initialize parameters, generate a global collaborative reconnaissance path, and allocate it to each reconnaissance UAV. The collaborative survey and real-time identification module is used to control the drone swarm to fly along a global collaborative survey path, simultaneously collect multi-source ground image data, and use an airborne AI model to identify snail breeding ground targets and their geographical coordinates in real time. The data fusion and task allocation module is used to receive all identification results, perform data fusion and deduplication, generate an electronic map of mollusc extermination target points, and dynamically allocate pesticide application tasks to the optimal pesticide application drone. The precision-targeted spraying module is used to control the pesticide application drone to accurately locate and operate the spraying mechanism to spray pesticides onto the target in an instantaneous and quantitative manner. The effect verification and report generation module is used to assign a survey drone to conduct a secondary scan after the application of pesticides. By comparing the changes in images before and after the application of pesticides using AI, the module can automatically verify the snail eradication effect and generate an operation report.

[0012] Beneficial effects of the present invention This invention provides a precise targeted snail eradication method and system based on UAV swarm collaboration and AI recognition. The method includes: receiving and parsing operational area information through a cloud-based control center, automatically generating a global collaborative reconnaissance path and allocating it to a reconnaissance UAV swarm; the UAV swarm synchronously collecting multi-source ground image data along the path, and using an onboard AI model to identify snail breeding ground targets and their geographical coordinates in real time; the cloud-based system aggregates all identification results, employs an improved DBSCAN algorithm for spatiotemporal clustering and data fusion to generate an electronic map of snail-eradicating targets, and dynamically allocates pesticide application tasks; the pesticide application UAV uses RTK positioning for centimeter-level precise positioning, controlling the spraying mechanism to spray a precise amount of pesticide onto the target center instantaneously; finally, the snail eradication effect is automatically verified through secondary scanning and AI image comparison. This invention automates the entire process of reconnaissance, identification, decision-making, spraying, and verification, greatly improving snail eradication efficiency and accuracy, significantly reducing pesticide usage, and eliminating the risk of personnel infection. Attached Figure Description

[0013] Figure 1This is a flowchart of a preferred embodiment of the present invention for a precise targeted snail eradication method based on drone swarm collaboration and AI recognition; Figure 2 This is a structural diagram of a precision-targeted snail eradication system based on drone swarm collaboration and AI recognition, according to a preferred embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. For ease of explanation, only the parts related to the embodiments of this invention are shown. It should be understood that the specific embodiments described herein are merely for explaining this invention and are not intended to limit this invention.

[0015] This invention provides a precise targeted snail eradication method and system based on UAV swarm collaboration and AI recognition. The method includes: receiving and parsing operational area information through a cloud-based control center, automatically generating a global collaborative reconnaissance path and allocating it to a reconnaissance UAV swarm; the UAV swarm synchronously collecting multi-source ground image data along the path, and using an onboard AI model to identify snail breeding ground targets and their geographical coordinates in real time; the cloud-based system aggregates all identification results, employs an improved DBSCAN algorithm for spatiotemporal clustering and data fusion to generate an electronic map of snail-eradicating targets, and dynamically allocates pesticide application tasks; the pesticide application UAV uses RTK positioning for centimeter-level precise positioning, controlling the spraying mechanism to spray a precise amount of pesticide onto the target center instantaneously; finally, the snail eradication effect is automatically verified through secondary scanning and AI image comparison. This invention achieves full-process automation of reconnaissance, identification, decision-making, spraying, and verification, greatly improving snail eradication efficiency and accuracy, significantly reducing pesticide usage, and eliminating the risk of personnel infection.

[0016] Figure 1 This is a flowchart of a preferred embodiment of the present invention for a precise targeted snail eradication method based on drone swarm collaboration and AI recognition; the method includes the following steps: S1, Receive and parse the work area information and initialize the parameters, generate a global collaborative reconnaissance path, and assign it to each reconnaissance UAV; including the following steps (steps S11-S13): Step S11: Input processing of work area information and parameter initialization; including (S111-S114): S111 receives the electronic map of the work area uploaded by the user (such as a KML / SHP format file), and parses the coordinates of its geographical boundary vertices; at the same time, it loads the digital elevation model (DEM) data layer corresponding to the area and the preset obstacle information layer (including no-fly zone polygons, high-voltage towers and other point and line elements and their safety buffers). S112, based on the snail eradication reconnaissance mission, initializes key flight and data acquisition parameters, including target reconnaissance resolution, image overlap rate, and safety interval; S113. Based on the scale of the snail breeding environment characteristics to be identified, determine the target ground sampling distance (GSD, e.g., 3 cm / pixel), and then calculate the required flight altitude (H). S114, Set the lateral overlap rate (O) for image acquisition. ≥ 60%) and heading overlap rate (O x ≥ 80%) Set the minimum safe vertical distance of the flight path relative to the ground and vegetation canopy; Step S12: Generate a global collaborative reconnaissance path. Using a regional coverage path planning algorithm, automatically generate a flight path that covers the entire operation area without repetition or omission. This includes (S121-S122):

[0017] S121, uses the ox-plowing reciprocating scanning method to generate the basic path; This involves using a plow-like reciprocating scanning method to generate a parallel reciprocating scanning path along the longest side of the circumscribed rectangle of the work area, in order to achieve the highest flight efficiency.

[0018] S122, perform basic path constraint processing and optimization; including... Spatial conflict detection is performed between the generated base path and the obstacle data layer; for path segments that intersect with no-fly zones or obstacle buffer zones, the A* algorithm (A-Star Algorithm) is used for local replanning to generate smooth, conflict-free detour paths. The coverage width (W) of a single image is calculated based on the sensor field of view (FOV) and flight altitude (H) of the reconnaissance drone; the coverage width (W) of a single image is calculated based on the preset lateral overlap ratio (O). The optimal strip spacing (S) is calculated using the formula: S = W × (1 - O) Meanwhile, based on the heading overlap ratio (O) x Calculate the waypoint spacing (L);

[0019] Step S13: The process of decomposing cluster tasks and allocating space includes (S131-S135): S131 employs an improved Veno diagram algorithm based on load balancing to divide the optimized global path into N sub-region paths (N being the number of reconnaissance drones). The key load indicator is set as the estimated reconnaissance time.

[0020] The estimated reconnaissance time is determined by the sub-region path length, number of turns, and terrain complexity, aiming to achieve a substantial balance in the workload of each UAV mission.

[0021] S132, preset different flight altitude layers for each sub-region path (for example, assign an altitude H_i = H_base + (i-1) * ΔH to the i-th drone, where H_base is the base flight altitude, which is determined by mission requirements; ΔH is the safe altitude interval, which is related to flight safety; and i is the index number of the drone); create a three-dimensional geofence for each sub-region to strictly limit the drone's activity range to its assigned area.

[0022] S133 generates an independent flight mission package for each sub-region path, which includes serialized waypoint coordinates, flight altitude, flight speed, camera trigger commands, and safety fence boundaries. S134 establishes a connection with all reconnaissance drones via the communication network to verify their GPS signal strength, battery level, and sensor status. After confirming that all units are in normal condition, the corresponding task packages are distributed to each reconnaissance drone.

[0023] S135 instructs the UAVs to automatically take off in sequence after receiving the mission package, autonomously navigate to the starting point of their respective assigned routes, enter standby status, and prepare to carry out collaborative reconnaissance missions.

[0024] S2, instructing the reconnaissance drone swarm to fly along the global collaborative reconnaissance path, simultaneously collecting multi-source raw ground image data, and performing real-time analysis based on a pre-trained airborne AI model to identify the target habitat of Oncomelania snails and its geographical coordinates; including the following steps (steps S21-S25): Step S21: Instruct the reconnaissance drone to fly synchronously along the assigned path and collect multi-source raw image data; In this embodiment of the invention, a high-definition zoom camera is used to acquire visible light images of the ground at a preset frequency (e.g., 1-2Hz); a multispectral or hyperspectral imager is used to acquire non-visible light band spectral data of the corresponding area; the spectral data includes environmental characteristics such as vegetation health status and moisture content; Step S22 involves real-time preprocessing of the acquired raw image data in the onboard computing unit, including: Convert the image into the input format required by the model; Image distortion correction and brightness normalization are performed to eliminate the influence of sensor and lighting conditions. Step S23: Input the preprocessed original image into the pre-trained airborne AI model for recognition processing and output Oncomelania snail breeding ground target information; the target information data includes the target's bounding box information and its corresponding confidence score; The pre-trained airborne AI model is a deep learning-based object detection network (preferably YOLO or Faster R-CNN architecture), and its processing includes the following steps: Deep features are extracted from images using a convolutional neural network backbone. Based on the aforementioned features, suspected Oncomelania snail breeding grounds (such as damp mudflats, specific vegetation communities, etc.) in image frames are identified and located. Output target information of Oncomelania snail breeding grounds, including the bounding box information of the target and its corresponding confidence score; Step S24: Bind target geographic information; including For each identified target, its pixel coordinates in the image are spatiotemporally aligned and fused with the latitude, longitude, and altitude obtained by the UAV during shooting via a high-precision positioning module (such as RTK-GPS), as well as the flight attitude (pitch, roll, yaw) data obtained via an inertial measurement unit (IMU). Based on camera sensor parameters and UAV pose data, the precise geodetic coordinates (longitude and latitude) of the target center point are calculated through forward intersection or projection transformation models. Step S25: Encapsulate and upload the target information in real time; including The identification result data is encapsulated into a structured target information data packet; the target information data packet includes at least: timestamp, target geodetic coordinates, identification confidence level, target area image snapshot, and UAV unique identifier; The target information data packets are uploaded to the cloud control center in real time and asynchronously via the airborne communication module and communication network, and then proceed to the next step of data fusion and decision-making process. S3 receives all identified target information data, performs data fusion and deduplication; generates a global electronic map of mollusc eradication targets; and dynamically allocates application tasks to the optimal application drone. This includes the following steps (steps S31-S35): Step S31: Multi-source data collection and standardization processing; including... The system continuously receives target information data uploaded by all reconnaissance drones via the communication network. The received target information data is subjected to integrity verification, and key fields such as timestamp, geographic coordinates, confidence level, and image evidence are parsed. All target information data are converted to a unified spatiotemporal coordinate system; Step S32: Perform spatiotemporal clustering and data fusion; including (S321-S323): S321, Establish a composite index structure based on geospatial and temporal dimensions; S322 uses the advanced DBSCAN algorithm for spatiotemporal clustering; Wherein: the spatial neighborhood threshold is set to ε=5 meters; the time window is set to Δt=10 minutes; and the minimum number of clustered samples is set to min_samples=2. Let there be two data points p and q, each with spatial coordinates... and timestamp Define the spatiotemporal composite distance function. for:

[0025] ; in, It is the spatial Euclidean distance; It is the absolute time difference. Δt is the time window threshold, set to 10 minutes; the ε-neighborhood Nϵ(p) of point p is defined as:

[0026] Nϵ(p)={q∈D∣dspace(p,q)≤ϵ and dtime(p,q)≤Δt}Nϵ; Where ϵ=5 meters is the spatial neighborhood threshold; The core algorithm for density clustering is: When the ε-neighborhood of point p contains at least min_samples=2 data points, p is identified as a core object; based on density reachability, points that are directly density reachable and density reachable are included in the same cluster; isolated points that cannot be included in any cluster are marked as noise data; S323, performing target deduplication and fusion: including Calculate the weighted center coordinates for each cluster; in, Pcenter This represents the center coordinates of the unique snail-killing target point generated after fusion. n This indicates the number of original identification results within the same target cluster; i This indicates the index number (from 1 to n) of the original recognition result currently being processed within the cluster. Pi Indicates the first i Geographic coordinates of the original identification results; with Indicates the first i The confidence level of each original identification result; Calculate the fusion confidence score for each cluster; in, Cfused This represents the overall confidence level of the unique molluscicide target after fusion. n This indicates the number of original recognition results. i Indicates the index number. Ci Indicates the first i The confidence level of each original identification result.

[0027] For each cluster, select the target image with the highest confidence level; Step S33: Generate an electronic map of snail eradication target sites; including (S331-S333): S331, add the merged target points as independent elements to the electronic map layer; S332, establishes a complete attribute record for each target point; The attribute records include: target unique identifier, precise geographic coordinates, fusion confidence score, associated image evidence index, and target level assessment (based on confidence score). S333, generate interactive electronic map services; The electronic map service supports target spatial query and filtering, attribute information visualization, map export and sharing; Step S34: Perform dynamic task allocation; including (S341-S344): S341, to obtain the real-time operating status of the spraying drone swarm; The operating status includes location coordinates and movement trajectory, remaining drug capacity, battery life, and current task load; S342, Establish an optimization model, that is, construct an optimization function with the objective of minimizing the total operation time; The model constraints include each target point being served by only one UAV, UAV drug capacity constraints, UAV endurance constraints, and mission completion time requirements; the optimization function is: min ∑ {i=1}^{m} ∑ {j=1}^{n} c_{ij} x_{ij} Where m represents the total number of spraying drones; n represents the total number of mollusc control targets to be sprayed; i represents the index of the spraying drone, i from 1 to m; j represents the index of the mollusc control target, j from 1 to n; c_{ij} represents the cost of the i-th drone traveling to the j-th target for operation. The cost can be flight distance, flight time, energy consumption, etc., or a combination of factors (such as distance, priority, etc.). In practical models, the goal is usually to minimize the total cost. x_{ij} is a decision variable, usually a 0-1 variable; if the i-th drone is assigned to spray the j-th target, then x_{ij} = 1; otherwise, it is 0.

[0028] S343 employs an improved genetic algorithm to optimize task allocation and obtain the optimal task allocation scheme; including... An integer encoding scheme is used to represent the task allocation strategy. A chromosome sequence of length n (total number of targets) is constructed, in which the value of each gene position represents the UAV number assigned to the corresponding target, forming a complete task allocation solution space representation.

[0029] Establish a fitness evaluation function that comprehensively considers the total path length, resource consumption balance, and task completion time balance; its expression is: F = α ⋅ D total+ β ⋅Δ R + c ⋅Δ T + d ⋅ P in: D total represents the total flight path length of the drone swarm; Δ R The range of liquid usage rates for each drone; Δ T The range of completion times for each drone mission; P To constrain the penalty terms for violations; for solutions that do not satisfy the constraints (such as exceeding the limits of liquid capacity, exceeding the limits of battery life, etc.), P takes a very large positive number; for feasible solutions, P=0. α , β , c , d These are the weighting coefficients.

[0030] Based on the above encoding and evaluation functions, selection, crossover, and mutation operations are performed to obtain an approximately optimal task allocation scheme through iterative evolution.

[0031] S344, Generate task instructions; including Generate the optimal operation sequence for each drone; calculate the optimal spraying parameters for each target point; generate a security check code containing path planning and spraying instructions; Step S35: Issue and confirm the task; including The system packages mission instructions into a standard communication protocol format; sends mission instructions to designated application drones via encrypted channels; receives mission confirmation signals from drones and updates the system mission status.

[0032] S4 instructs the assigned pesticide application drone to accurately locate itself using the precision positioning module and controls the spraying mechanism to spray pesticide onto the target center instantaneously and quantitatively; the precision positioning module is an RTK-GPS module, achieving centimeter-level positioning accuracy; The spraying mechanism is a nozzle or piezoelectric nozzle controlled by a miniature solenoid valve, which can achieve pulse spraying with millisecond-level start and stop; Navigation and Precise Positioning: After receiving instructions, the pesticide-dispensing drone flies to its first assigned target coordinates. As it approaches the target, it relies on its onboard RTK-GPS module to improve positioning accuracy from meters to centimeters (typically ±1-3 centimeters), thereby achieving stable and precise hovering over the target.

[0033] Instantaneous quantitative spraying: Start-stop control: The flight control system sends pulse signals to the precision spraying mechanism (such as a miniature high-speed solenoid valve). The solenoid valve has a response time in the millisecond range, enabling instantaneous opening and closing of the spray solution.

[0034] Quantitative spraying: By controlling the opening time of the nozzle (e.g., 0.2-1.0 seconds) and the liquid pressure, ensure that the dosage of the agent sprayed each time is a pre-calibrated fixed small dose (e.g., 5-10 ml).

[0035] S5. After the pesticide application is completed, a reconnaissance drone is assigned to conduct a second scan of the sprayed area. AI is used to compare the changes in images before and after application, automatically verifying the snail eradication effect and generating an operation report. This includes (steps S51-S53):

[0036] Step S51: Assign an idle or optimally positioned reconnaissance drone to perform a secondary flight scan of the sprayed target area and collect post-application image data.

[0037] Step S52: The acquired post-processed image of the target and the stored original image of the target are both fed into the AI ​​model for comparison. The comparison method can be:

[0038] Change detection: Using image difference or semantic segmentation models, identify whether the color, texture and other features of the target area have changed significantly due to the drug effect (such as vegetation withering or the ground color darkening).

[0039] Activity assessment: The target detection model directly determines whether the original breeding environment characteristics still exist or whether their confidence level has decreased significantly.

[0040] Step S53: Summarize all task data and generate a structured job report.

[0041] The report includes, but is not limited to: total operating area, total flight mileage, number of original targets identified, number of targets after fusion, number of targets sprayed, total amount of pesticide consumed, average amount of pesticide consumed per target, number of sampling points for effect verification, and snail eradication efficiency.

[0042] Corresponding to the above embodiment of a precise targeted snail eradication method based on drone swarm collaboration and AI recognition, Figure 2 This diagram illustrates a structural block diagram of a precision-targeted snail eradication system based on UAV swarm collaboration and AI recognition, according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The system includes:

[0043] The task planning and cluster initialization module is used to receive work area information, parse and initialize parameters, generate a global collaborative reconnaissance path, and allocate it to each reconnaissance UAV. The collaborative survey and real-time identification module is used to control the drone swarm to fly along a global collaborative survey path, simultaneously collect multi-source ground image data, and use an airborne AI model to identify snail breeding ground targets and their geographical coordinates in real time. The data fusion and task allocation module is used to receive all identification results, perform data fusion and deduplication, generate an electronic map of mollusc extermination target points, and dynamically allocate pesticide application tasks to the optimal pesticide application drone. The precision-targeted spraying module is used to control the pesticide application drone to accurately locate and operate the spraying mechanism to spray pesticides onto the target in an instantaneous and quantitative manner. The effect verification and report generation module is used to assign a survey drone to conduct a secondary scan after the application of pesticides. By comparing the changes in images before and after the application of pesticides using AI, the module can automatically verify the snail eradication effect and generate an operation report.

[0044] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by program instructions and related hardware. The program can be stored in a computer-readable storage medium, such as ROM, RAM, disk, optical disk, etc.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A precise target snail elimination method based on UAV swarm cooperation and AI identification, characterized in that, The method comprises the following steps: Receiving and analyzing the work area information and initializing parameters, generating a global collaborative survey path, and assigning it to each survey unmanned aerial vehicle; Instructing the survey unmanned aerial vehicle group to fly along the global collaborative survey path, synchronously collecting multi-source raw image data on the ground, and performing real-time analysis based on a pre-trained AI model to identify snail breeding ground targets and their geographic coordinates; Receiving all identified target information data, performing data fusion and deduplication; Generating a global snail eradication target electronic map; And dynamically assigning the drug application task to the optimal drug application unmanned aerial vehicle; The assigned drug application unmanned aerial vehicle performs precise positioning and controls the spraying mechanism to perform instantaneous and quantitative drug injection on the target center. After the drug application is completed, the survey unmanned aerial vehicle is instructed to perform secondary scanning on the sprayed area, automatically verifies the snail eradication effect by comparing the image changes before and after the drug application, and generates a work report.

2. The precision targeting and killing method based on UAV swarm cooperation and AI identification according to claim 1, wherein Receiving and analyzing the work area information and initializing parameters, including the following steps: Receiving the user-uploaded work area electronic map and analyzing its geographic boundary vertex coordinates; at the same time, loading the digital elevation model data layer and the preset obstacle information layer corresponding to the area; According to the snail eradication survey task, initializing the key flight and collection parameters, including the target survey resolution, image overlap rate, and safety interval; According to the required identification of the snail breeding environment feature scale, the target ground sampling distance is determined, and the required flight height is calculated; Set the lateral overlap rate and heading overlap rate of image collection; Set the minimum safe vertical distance of the flight path relative to the ground and the vegetation canopy.

3. The precision targeting snail elimination method based on UAV swarm cooperation and AI identification of claim 1, wherein The global collaborative survey path is generated using a regional coverage path planning algorithm, which includes: Using the ploughing type reciprocating scanning method, parallel reciprocating scanning paths are generated along the longest side direction of the circumscribed rectangle of the work area to achieve the highest flight efficiency; The generated basic path is subjected to spatial conflict detection with the obstacle data layer. For path segments intersecting with no-fly zones or obstacle buffer zones, the A* algorithm is used for local re-planning to generate smooth conflict-free detour paths; According to the field angle of view of the survey unmanned aerial vehicle and the flight height, the coverage width of a single image is calculated; according to the preset lateral overlap rate, the optimal strip spacing is calculated.

4. The precision target snail eradication method based on UAV swarm cooperation and AI identification of claim 1, wherein The cluster task decomposition and airspace allocation includes Using an improved Voronoi diagram algorithm based on load balancing, the optimized global path is divided into a certain number of sub-regional paths; The load key indicator is set as the estimated survey time; Different flight altitudes are preset for each sub-regional path; a three-dimensional geographic fence is created for each sub-region to strictly limit the activity range of the unmanned aerial vehicle within its assigned area; An independent flight task package is generated for each sub-regional path; Through a communication network, all survey unmanned aerial vehicles are connected to check their GPS signal strength, battery capacity, and sensor status; after confirming that all unit states are normal, the corresponding task package is respectively sent to each survey unmanned aerial vehicle; After receiving the task package, the unmanned aerial vehicle automatically takes off in turn and autonomously navigates to the starting point of the assigned flight path, enters the standby state, and is ready to perform the collaborative survey task.

5. The precision targeting and killing method of oncomelania based on UAV swarm cooperation and AI identification according to claim 1, wherein The indication survey unmanned aerial vehicle group flies along the global cooperative survey path, synchronously collects multi-source original image data on the ground, and performs real-time analysis based on a pre-trained on-board AI model to identify oncomelania breeding ground targets and their geographic coordinates, including: The indication survey unmanned aerial vehicle group flies along the global cooperative survey path, synchronously collects multi-source original image data on the ground, and performs real-time analysis based on a pre-trained on-board AI model to identify oncomelania breeding ground targets and their geographic coordinates, including: The indication survey unmanned aerial vehicle group flies along the global cooperative survey path, synchronously collects multi-source original image data on the ground, and performs real-time analysis based on a pre-trained on-board AI model to identify oncomelania breeding ground targets and their geographic coordinates, including: The indication survey unmanned aerial vehicle group flies along the global cooperative survey path, synchronously collects multi-source original image data on the ground, and performs real-time analysis based on a pre-trained on-board AI model to identify oncomelania breeding ground targets and their geographic coordinates, including: The indication survey unmanned aerial vehicle group flies along the global cooperative survey path, synchronously collects multi-source original image data on the ground, and performs real-time analysis based on a pre-trained on-board AI model to identify oncomelania breeding ground targets and their geographic coordinates, including: The pre-trained on-board AI model is a deep learning-based target detection network, and its processing includes the following steps:

6. The precision target snail eradication method based on UAV swarm cooperation and AI identification of claim 5, wherein Deep features in the image are extracted through a convolutional neural network backbone network; Based on the features, suspected oncomelania breeding grounds in the image frame are identified and located; Oncomelania breeding ground target information is output, including the target's bounding box information and its corresponding confidence score; The target geographic information binding includes For each identified target, its pixel coordinates in the image are spatiotemporally aligned and fused with the unmanned aerial vehicle's acquired latitude, longitude, and altitude, as well as the acquired flight attitude data; Based on the camera sensor parameters and the unmanned aerial vehicle's pose data, the precise geodetic coordinates of the target center point are calculated through forward intersection or projection transformation model. The received target information data is fused and de-duplicated; 7. The precision target snail eradication method based on UAV swarm cooperation and AI identification of claim 1, wherein, A global oncomelania eradication target electronic map is generated; And the optimal oncomelania eradication unmanned aerial vehicle is dynamically assigned a pesticide application task, including: Multi-source data collection and standardization processing; Spatiotemporal clustering and data fusion are performed; An oncomelania eradication target electronic map is generated; Dynamic task allocation is performed; Task assignment and confirmation are performed. The spatiotemporal clustering and data fusion include 8. The precision target snail eradication method based on UAV swarm cooperation and AI identification of claim 7, wherein A composite index structure based on geographic space and time dimension is established; The DBSCAN algorithm is used for spatiotemporal clustering; Target de-duplication and fusion are performed. The dynamic task allocation includes 9. The precision target snail eradication method based on UAV swarm cooperation and AI identification of claim 7, wherein The running state of the oncomelania eradication unmanned aerial vehicle group is acquired in real time, including position coordinates and motion trajectory, remaining liquid capacity, battery endurance time, and current task load; An optimization model is established, i.e., an optimization function is constructed with the goal of minimizing total operation time; An improved genetic algorithm is used for allocation optimization solution to obtain the optimal task allocation scheme; Task instructions are generated, including generating an optimal operation sequence for each unmanned aerial vehicle and calculating the best spraying parameters for each target; A safety check code containing path planning and spraying instructions is generated; The improved genetic algorithm is used for allocation optimization solution to obtain the optimal task allocation scheme, including: A chromosome sequence with a length equal to the total number of targets is constructed, where each gene site value represents the number of the unmanned aerial vehicle assigned to the corresponding target, forming a complete task allocation solution space representation; An adaptability evaluation function is established, which comprehensively considers the total path length, resource consumption balance, and task completion time balance, and the adaptability evaluation function expression is: The system includes: F = α ⋅ D total+ β ⋅Δ R + ​ ⋅Δ T + ​ ⋅ P ; wherein: D total is the total flight path length of the UAV fleet; Δ R is the range of the liquid usage rate of each UAV; Δ T is the range of the task completion time of each UAV; P is the constraint violation penalty term; for a solution that does not satisfy the constraints, P takes a large positive number, and for a feasible solution, P = 0; α , β , ​ , ​ is the weight coefficient.

10. A precise target snail elimination system based on UAV swarm cooperation and AI identification, characterized in that, ​ The task planning and cluster initialization module is configured to receive job area information, analyze and initialize parameters, generate a global cooperative survey path, and distribute the global cooperative survey path to each survey unmanned aerial vehicle; The cooperative survey and real-time identification module is configured to control the unmanned aerial vehicle group to fly according to the global cooperative survey path, synchronously collect multi-source image data of the ground, and use an on-board AI model to identify a snail breeding ground target and a geographic coordinate of the snail breeding ground target in real time; The data fusion and task distribution module is configured to receive all identification results, perform data fusion and deduplication, generate an electronic map of a snail elimination target, and dynamically distribute a pesticide application task to an optimal pesticide application unmanned aerial vehicle; The precise target spraying module is configured to control the pesticide application unmanned aerial vehicle to perform precise positioning, and operate a spraying mechanism to perform instantaneous and quantitative pesticide spraying on a target center; The effect verification and report generation module is configured to assign a survey unmanned aerial vehicle to perform secondary scanning after the pesticide application is completed, automatically verify a snail elimination effect by comparing image changes before and after the pesticide application, and generate a job report.

Images