Unmanned aerial vehicle swarm cooperative environment detection and pollution source tracing method and system for complex terrain

Abstract

This invention provides a method and system for collaborative environmental monitoring and pollution source tracing using UAV swarms in complex terrain, relating to the fields of UAV environmental monitoring and pollution source tracing. The method includes: a ground control center planning layered and altitude-based tasks for the target area and sending them to each group containing master and sub-UAVs; the master UAV decomposing and encoding the sub-tasks and transmitting the results back; the ground-based optimization of flight paths being distributed; the sub-UAVs collecting pollution source data, which is then preprocessed by the master UAV and uploaded; ground-based analysis results being fed back; and the master UAV calculating the contribution value C using a source tracing model. en Once the source tracing is complete, the results are uploaded to the server, and the drone returns to base. The system includes a ground control center and at least one drone swarm; each swarm consists of a master drone and sub-drones. This invention can plan flight paths based on the geographical environment of the monitored target area, avoiding detection blind spots and conflicts, and coordinate actions based on the flight path, thus improving the convenience and accuracy of environmental monitoring and pollution source tracing.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) environmental monitoring and pollution source tracing, and in particular to a method and system for collaborative environmental monitoring and pollution source tracing using UAV swarms in complex terrain. Background Technology

[0002] Currently, environmental pollution has become a pressing global challenge, with pollutants exceeding standards in key environmental media such as the atmosphere, water bodies, and soil. The prerequisite for effective pollution control is the accurate tracing and location of pollution sources, which requires monitoring systems to possess comprehensive capabilities including wide coverage, high-precision detection, and efficient source tracing. However, existing mainstream monitoring technologies all have significant shortcomings: ground-based monitoring equipment not only consumes a lot of energy and is limited by line-of-sight range, but also has low spatial resolution, making it difficult to accurately locate small-scale pollution sources; while orbital satellites can cover a wide area, their detection accuracy is insufficient, and they are constrained by time and orbit, making it impossible to capture sudden pollution events in a timely manner. Furthermore, existing technologies mostly focus on the detection of multi-component pollutants or the development of single source tracing models, lacking an efficient integrated "detection-source tracing" process. This makes it difficult to balance coverage, efficiency, and accuracy, becoming a core bottleneck restricting the effectiveness of environmental monitoring.

[0003] With its significant advantages of high flexibility, ease of operation, and controllable cost, unmanned aerial vehicle (UAV) technology can effectively compensate for the shortcomings of ground-based and satellite monitoring, providing a new technological path to solve the aforementioned problems. However, it should be noted that the flight endurance and payload capacity of a single UAV are limited, making it difficult to meet the actual needs of large-scale pollutant detection and complex pollution source tracing. Therefore, collaborative monitoring by multiple UAVs has become an inevitable trend for future development. Nevertheless, current related technologies still have significant shortcomings: on the one hand, the mission planning and pollutant detection of a single UAV largely rely on traditional flight control systems and sensors, and there is still considerable room for improvement in mission complexity and detection precision; on the other hand, the industry has not yet established an efficient collaborative pollutant detection and source tracing process, and existing technologies are clearly fragmented—some technologies can only complete the detection of multiple pollutant concentrations, providing only information on the spatial distribution of pollution, lacking source tracing analysis capabilities; some technologies, while focusing on source tracing analysis, still rely on the pre-detection of single or multi-component pollutants, resulting in inefficient process connections; and some technologies directly conduct pollution source tracing but lack data support from pollutant component detection, making it difficult to guarantee the accuracy of source tracing.

[0004] In summary, there is still significant room for improvement in current drone-based pollution monitoring and source tracing technologies. How to fully leverage the flexibility and low cost advantages of drones to overcome the collaborative technical bottlenecks of "wide-area coverage—multi-component detection—high-precision source tracing" and build an integrated and efficient process is not only crucial for filling existing technological gaps but also holds significant practical importance for enhancing the targeting of pollution control and strengthening health risk early warning capabilities. This is a key issue that urgently needs to be addressed. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides a method and system for collaborative environmental monitoring and pollution source tracing using unmanned aerial vehicle (UAV) swarms in complex terrain. It can plan flight paths based on the geographical environment of the target monitoring area, avoid detection blind spots and detection conflicts, and coordinate according to the flight paths, thereby improving the convenience and accuracy of environmental monitoring and pollution source tracing.

[0006] The technical solution adopted in this invention is:

[0007] A method for collaborative environmental monitoring and pollution source tracing using UAV swarms in complex terrain includes the following steps:

[0008] Step S1: The ground control center pre-plans the hierarchical and height-based tasks for the target area and sends them to each corresponding UAV group; each UAV group contains one master UAV and the rest are sub-UAVs.

[0009] In step S2, the main UAV of the corresponding group receives the hierarchical and altitude-divided task, decomposes it into multiple sub-tasks, and then encodes the multiple sub-tasks according to the agreed encoding rules to form encoded data, which is then sent to the ground control center.

[0010] Step S3: The ground control center receives the coded data, optimizes the mission data transmission, generates the planned flight path for each sub-mission, and then sends the planned flight path to the master UAV of the corresponding group.

[0011] In step S4, the main UAV of the corresponding group receives the planned flight path, combines the hierarchical and altitude-based tasks, performs task allocation and dynamic trajectory planning, so that the sub-UAVs in the same group collect pollution source-related data through the onboard pollutant detection module according to the assigned tasks and execution routes, and upload it to the main UAV in real time for preprocessing to form complete pollution source information, and send it to the ground control center.

[0012] In step S5, the ground control center receives pollution source information, performs pollutant component analysis, forms pollutant analysis conclusions, and then sends them to the main UAV of the corresponding group.

[0013] Step S6: The main UAV of the corresponding group receives the pollutant analysis results, combines them with the flight location information, and uses the built-in pollution source tracing model to calculate the contribution value C of each pollutant source to the pollutant source through a quantitative formula. en The system calculates and traces the intensity of pollution sources; after tracing the pollution sources, it sends the results to the ground control center; the drone returns, and the ground control center uploads the pollutant detection and tracing results to the server.

[0014] Furthermore, the layered and altitude-based tasks in step S1 include flight altitude layering and monitoring task layering: the flight altitude is divided into at least three layers from low to high, with different altitude layers corresponding to different monitoring tasks; among them, the low-level monitoring task is to screen the concentration of major pollutants including nitrogen oxides and sulfur-containing compounds, the mid-level monitoring task is to accurately detect the concentration of volatile organic compounds and particulate matter, and the high-level monitoring task is to detect nitrogen dioxide and carbon monoxide with high sensitivity.

[0015] Furthermore, the specific execution process in step S3 is as follows:

[0016] Step S31: After receiving the coded data, the ground control center parses it according to the coding rules corresponding to the monitoring tasks under different environments to obtain multiple sub-tasks;

[0017] Step S32: Calculate the overlap between any two subtasks, and merge subtasks whose overlap meets the preset conditions according to the task distribution under different environments.

[0018] Step S33: Generate the planned flight paths of each subtask after merging in different scenarios on a two-dimensional plane using the KDTree algorithm, and then send the planned flight paths to the master UAV of the corresponding group.

[0019] Furthermore, the specific execution process in step S32 is as follows:

[0020] Step S321: Extract the core feature parameters of each subtask. The core feature parameters include spatial region parameters (such as latitude and longitude range) representing the boundary coordinates of the monitoring area corresponding to the subtask, pollutant type parameters (such as sulfur dioxide and nitrogen oxides) representing the set of pollutant types to be monitored by the subtask, and terrain adaptation parameters (such as values ​​of 0-1 based on altitude difference and slope preset, with higher values ​​indicating more complex terrain) representing the terrain complexity coefficient of the subtask area.

[0021] Step S322: Based on spatial region parameters, pollutant type parameters, and terrain adaptation parameters, calculate the spatial overlap degree R1, pollutant type overlap degree R2, and terrain adaptation overlap degree R3, respectively; where,

[0022] Spatial overlap R1 = Intersection area of ​​the two sub-task monitoring areas / Union area of ​​the two sub-task monitoring areas;

[0023] Pollutant type overlap R2 = Number of pollutant types jointly monitored by the two sub-tasks / Total number of pollutant types monitored by the two sub-tasks after deduplication;

[0024] Terrain adaptation overlap R3: When the difference in terrain complexity coefficient between two subtasks is less than or equal to the first preset threshold (e.g., 0.2), R3 = 1; otherwise, R3 = 0.

[0025] Step S323: Calculate the overall overlap R = α·R1 + β·R2 + γ·R3; where α, β, and γ are the weight coefficients of spatial region parameters, pollutant type parameters, and terrain adaptation parameters, respectively, and α + β + γ = 1. The weights are dynamically adjusted according to environmental characteristics. For complex terrain areas (such as mountains / canyons), the weight of α is increased to 0.5, and for areas with dense pollutants (such as industrial areas), the weight of β is increased to 0.5.

[0026] Step S324: If the overall overlap R is greater than or equal to the second preset threshold (e.g., 0.6), then the two subtasks are determined to meet the overlap condition and are merged.

[0027] Furthermore, the specific execution process of the trajectory dynamic planning in step S4 is as follows:

[0028] Step S41: The main UAV acquires digital elevation model data of the planned flight path and target area, and extracts the coordinates and elevations of terrain feature points (such as peaks, valleys, and steep slopes).

[0029] Step S42: Construct a three-dimensional obstacle avoidance constraint model based on digital elevation model data, using the minimum safe flight altitude, turning radius, and remaining energy consumption of the sub-UAV as constraints.

[0030] Step S43: A heuristic search algorithm is used to introduce an energy consumption cost factor to perform three-dimensional correction on the planned flight path, generating a sub-drone execution route with "obstacle avoidance + optimal energy consumption"; where, the energy consumption cost factor = the current remaining power of the sub-drone / the estimated energy consumption to complete the route segment, and priority is given to allocating short-path, low-altitude and non-steep-slope execution routes to sub-drones with low power.

[0031] Step S44: If the sub-UAV detects a sudden terrain obstacle (such as a collapsed body or an unmarked steep slope) during flight, the main UAV receives the obstacle coordinates in real time, re-executes steps S41 to S43, generates a temporary alternative route, and sends it to the corresponding sub-UAV.

[0032] Furthermore, the process of constructing the pollution source tracing model in step S6 is as follows:

[0033] The flight locations and pollutant components (such as nitrogen oxides and nitrogen dioxide) of historical sub-tasks are used as inputs to the convolutional neural network, and the pollution source tracing results of historical sub-tasks are used as training labels.

[0034] By training a convolutional neural network with training labels, a pollution source tracing model can be obtained that can trace pollution sources in different environments.

[0035] Furthermore, the specific execution process of pollution source intensity calculation and source tracing in step S6 is as follows:

[0036] Step S61: The main UAV integrates the pollutant analysis conclusions, the flight location information of the sub-UAVs (such as latitude and longitude, altitude) and the basic parameters of the monitoring points (such as terrain slope, ambient temperature and humidity, wind speed), and generates a standardized source tracing calculation dataset after removing missing values ​​and outliers.

[0037] Step S62: Input the source tracing calculation dataset into the built-in pollution source tracing model based on convolutional neural network, and output the candidate pollution source list and preliminary weights for each pollutant at each monitoring point after feature extraction.

[0038] Step S63, according to formula C en =ω en ·S e +B n Calculate the contribution of candidate pollution sources to the corresponding pollutants; where C en Let ω be the contribution of the e-th pollution source to the n-th pollutant. en S represents the contribution ratio of the pollution source tracing model output based on pollutant transport patterns and complex terrain features. e B represents the pollutant emission intensity of the pollution source per unit time. n The background concentration of pollutants in the target area that is not subject to human pollution;

[0039] Step S64: For each pollutant at a single monitoring point, sort them from high to low according to their contribution value, screen out the top (e.g., the first) core pollution source, and combine the terrain features to clarify the transmission path of pollutants from the core pollution source to the monitoring point.

[0040] Step S65: Integrate the source tracing results of individual locations to generate a heat map of pollution source intensity distribution in the area covered by the drone and the overall source tracing conclusion.

[0041] Furthermore, after step S6, a multi-machine cluster traceability data cross-validation step is also included, the specific execution process of which is as follows:

[0042] Step S71: The ground control center collects the source tracing conclusions uploaded by all UAV groups and extracts the source tracing results of the same pollutant in the overlapping areas of the monitoring area from different UAV groups;

[0043] Step S72: Calculate the deviation rate of the pollution source contribution value of the same pollutant in the overlapping area. Deviation rate = |C calculated by Unit A en- Unit B's calculation of C en | / max(C calculated by unit A) en Unit B's calculation of C en ); where C en Let e ​​be the contribution of the e-th pollution source to the n-th pollutant;

[0044] Step S73: If the deviation rate is greater than the preset deviation threshold (e.g., 15%), then the high-resolution satellite remote sensing data and historical data from ground monitoring stations in the overlapping area are used as references, and the deviation source tracing results are calibrated by combining the Gaussian plume model corrected for complex terrain.

[0045] Step S74: Integrate the source tracing results of all calibrated units and regenerate the overall source tracing report for the target area.

[0046] Furthermore, in step S6, the pollution source tracing model undergoes dynamic iterative optimization during use, specifically including:

[0047] Step S661: The ground control center summarizes the actual detection data, source tracing results, and pollution source verification results of the UAV swarm on a weekly / monthly basis;

[0048] Step S662: Compare the verification results with the source tracing conclusions output by the model, and calculate the model source tracing accuracy rate; where, the model source tracing accuracy rate = number of core pollution sources confirmed by verification / number of core pollution sources output by the model;

[0049] Step S663: If the accuracy is less than the third preset threshold (e.g., 85%), then filter out samples with low accuracy (e.g., source data in canyon terrain or high humidity environment) and add them to the model training set.

[0050] Step S664: The original convolutional neural network model is fine-tuned using the transfer learning method (freezing the bottom feature extraction layer and updating only the top fully connected layer) to generate an iterative model adapted to the dynamic pollution features of complex terrain.

[0051] In step S665, the ground control center distributes the iterated model to all main UAVs, replacing the original pollution source tracing model and realizing the self-adaptive optimization of the model.

[0052] Based on the same inventive concept, the present invention also provides a UAV swarm collaborative environmental detection and pollution source tracing system for complex terrain, to implement the aforementioned UAV swarm collaborative environmental detection and pollution source tracing method for complex terrain, characterized in that it includes: a ground control center, and at least one UAV swarm connected to the ground control center; each UAV swarm includes a master UAV and several sub-UAVs.

[0053] The ground control center is equipped with:

[0054] The task planning module is used to pre-plan the hierarchical and altitude-based tasks for the target area and send the hierarchical and altitude-based tasks to each corresponding drone cluster.

[0055] The encoding data processing module is used to receive the encoded data uploaded by the main UAV, optimize the data transmission of the encoded data, generate the planned flight path for each sub-task, and send the planned flight path to the corresponding group of main UAVs.

[0056] The pollutant analysis module is used to receive complete pollution source information uploaded by the main UAV, analyze the pollutant components, form pollutant analysis conclusions, and send the pollutant analysis conclusions to the corresponding group's main UAV;

[0057] The results management module is used to receive the pollution source tracing results uploaded by the main drone. After the drone returns, it summarizes and uploads the pollutant detection and tracing results to the server.

[0058] The main drone is equipped with:

[0059] The subtask processing module is used to receive hierarchical and high-level tasks, decompose them into multiple subtasks, encode the multiple subtasks into encoded data according to the agreed encoding rules, and send them to the ground control center.

[0060] The trajectory planning module is used to receive planned flight paths, combine hierarchical and altitude-based tasks to complete task allocation and dynamic trajectory planning, and control the sub-UAVs in the same group to perform data collection according to the assigned tasks and execution routes.

[0061] The data preprocessing module is used to receive pollution source-related data uploaded by the sub-UAV, preprocess it, form complete pollution source information, and then send it to the ground control center.

[0062] The source tracing calculation module, with a built-in pollution source tracing model, receives pollutant analysis results and, combined with flight location information, calculates the contribution value C of each pollutant source to each pollutant using a quantitative formula. en It also completes the calculation and tracing of pollution source intensity, generates pollution source tracing results, and sends them to the ground control center;

[0063] The sub-drone is equipped with:

[0064] The pollutant detection module is mounted on a sub-drone to collect pollution source-related data according to the assigned tasks and execution routes, and uploads the pollution source-related data to the main drone in real time.

[0065] The beneficial effects of this invention are:

[0066] This invention discloses a method and system for collaborative environmental monitoring and pollution source tracing using unmanned aerial vehicle (UAV) swarms in complex terrain. It employs a master-slave UAV swarm architecture, hierarchical and altitude-based task planning, an integrated monitoring-tracing process, and a quantitative formula to calculate the pollution source contribution value C. enThis system connects the ground control center with the drone swarm for full-process collaboration, breaking through the limitations of traditional ground monitoring line-of-sight, insufficient satellite monitoring accuracy, and the coverage shortcomings of single drones. It achieves a collaborative closed loop of "wide coverage - multi-component detection - high-precision source tracing," with layered and height-based task planning to adapt to the three-dimensional monitoring needs of complex terrain. The master-slave architecture improves task execution efficiency, and the quantitative formula makes the source tracing results quantifiable. It solves the core bottleneck of existing technologies that "difficulty in balancing range, efficiency, and accuracy," providing precise data support for pollution control. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a flowchart illustrating the method for collaborative environmental monitoring and pollution source tracing using a swarm of drones in complex terrain, as described in Example 1. Detailed Implementation

[0069] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0070] The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. To simplify the disclosure of the present invention, the components and arrangements of specific examples are described below. Of course, these are merely examples and are not intended to limit the present invention.

[0071] The embodiments of the invention will now be described in detail with reference to the accompanying drawings.

[0072] Example 1

[0073] This embodiment provides a method for collaborative environmental monitoring and pollution source tracing using a swarm of unmanned aerial vehicles (UAVs) in complex terrain. The process is shown in the attached figure. Figure 1 As shown in the figure. The method includes the following steps:

[0074] Step S1: The ground control center pre-plans the hierarchical and height-based tasks for the target area and sends them to each corresponding UAV group; each UAV group contains one master UAV and the rest are sub-UAVs.

[0075] In step S2, the main UAV of the corresponding group receives the hierarchical and altitude-divided task, decomposes it into multiple sub-tasks, and then encodes the multiple sub-tasks according to the agreed encoding rules to form encoded data, which is then sent to the ground control center.

[0076] Step S3: The ground control center receives the coded data, optimizes the mission data transmission, generates the planned flight path for each sub-mission, and then sends the planned flight path to the master UAV of the corresponding group.

[0077] In step S4, the main UAV of the corresponding group receives the planned flight path, combines the hierarchical and altitude-based tasks, performs task allocation and dynamic trajectory planning, so that the sub-UAVs in the same group collect pollution source-related data through the onboard pollutant detection module according to the assigned tasks and execution routes, and upload it to the main UAV in real time for preprocessing to form complete pollution source information, and send it to the ground control center.

[0078] In step S5, the ground control center receives pollution source information, performs pollutant component analysis, forms pollutant analysis conclusions, and then sends them to the main UAV of the corresponding group.

[0079] Step S6: The main UAV of the corresponding group receives the pollutant analysis results, combines them with the flight location information, and uses the built-in pollution source tracing model to calculate the contribution value C of each pollutant source to the pollutant source through a quantitative formula. en The system calculates and traces the intensity of pollution sources; after tracing the pollution sources, it sends the results to the ground control center; the drone returns, and the ground control center uploads the pollutant detection and tracing results to the server.

[0080] The beneficial effects of adopting the above technical solutions are: employing a "master-slave drone swarm architecture + hierarchical and hierarchical task planning + integrated detection-source tracing process + quantitative formula to calculate the pollution source contribution value C". en This system connects the ground control center with the drone swarm for full-process collaboration, breaking through the limitations of traditional ground monitoring line-of-sight, insufficient satellite monitoring accuracy, and the coverage shortcomings of single drones. It achieves a collaborative closed loop of "wide coverage - multi-component detection - high-precision source tracing," with layered and height-based task planning to adapt to the three-dimensional monitoring needs of complex terrain. The master-slave architecture improves task execution efficiency, and the quantitative formula makes the source tracing results quantifiable. It solves the core bottleneck of existing technologies that "difficulty in balancing range, efficiency, and accuracy," providing precise data support for pollution control.

[0081] In this further optimized embodiment, the layered and altitude-based task in step S1 includes flight altitude layering and monitoring task layering: the flight altitude is divided into at least three layers from low to high, with different altitude layers corresponding to different monitoring tasks. For example, according to altitude, it can be divided into a bottom layer of 0-100m, a middle layer of 100-300m, and a high layer of 300-500m. Among them, the monitoring task of the bottom layer is to screen the concentration of major pollutants including nitrogen oxides and sulfur-containing compounds; the monitoring task of the middle layer is to accurately detect the concentration of volatile organic compounds and particulate matter; and the monitoring task of the high layer is to detect nitrogen dioxide and carbon monoxide with high sensitivity.

[0082] The beneficial effects of adopting the above technical solutions are as follows: The hierarchical and altitude-based monitoring tasks are refined into "flight altitude stratification + differentiated monitoring tasks." Lower altitudes screen major pollutant categories, mid-levels accurately detect characteristic pollutants, and upper altitudes provide highly sensitive monitoring of strong pollutant components. Addressing the differences in pollutant distribution at different altitudes in complex terrain, this achieves hierarchical monitoring of pollutants through "major category screening - precise detection - key monitoring," avoiding the missed detections and false detections of traditional single-monitoring models. It also clarifies the monitoring priorities at each altitude level, ensuring both large-scale pollution coverage and improving the accuracy of characteristic pollutant detection. This solves the pain point of "inefficient integration of multi-component detection and source tracing" in existing technologies, providing targeted data for subsequent source tracing and improving tracing efficiency.

[0083] In the further optimized scheme of this embodiment, the specific execution process of step S3 is as follows:

[0084] Step S31: After receiving the coded data, the ground control center parses it according to the coding rules corresponding to the monitoring tasks under different environments to obtain multiple sub-tasks;

[0085] Step S32: Calculate the overlap between any two subtasks, and merge subtasks whose overlap meets the preset conditions according to the task distribution under different environments.

[0086] Step S33: Generate the planned flight paths of each subtask after merging in different scenarios on a two-dimensional plane using the KDTree algorithm, and then send the planned flight paths to the master UAV of the corresponding group.

[0087] The specific execution process in step S32 is as follows:

[0088] Step S321: Extract the core feature parameters of each subtask. The core feature parameters include spatial region parameters (such as latitude and longitude range) representing the boundary coordinates of the monitoring area corresponding to the subtask, pollutant type parameters (such as sulfur dioxide and nitrogen oxides) representing the set of pollutant types to be monitored by the subtask, and terrain adaptation parameters (such as values ​​of 0-1 based on altitude difference and slope preset, with higher values ​​indicating more complex terrain) representing the terrain complexity coefficient of the subtask area.

[0089] Step S322: Based on spatial region parameters, pollutant type parameters, and terrain adaptation parameters, calculate the spatial overlap degree R1, pollutant type overlap degree R2, and terrain adaptation overlap degree R3, respectively; where,

[0090] Spatial overlap R1 = Intersection area of ​​the two sub-task monitoring areas / Union area of ​​the two sub-task monitoring areas;

[0091] Pollutant type overlap R2 = Number of pollutant types jointly monitored by the two sub-tasks / Total number of pollutant types monitored by the two sub-tasks after deduplication;

[0092] Terrain adaptation overlap R3: When the difference in terrain complexity coefficient between two subtasks is less than or equal to the first preset threshold (e.g., 0.2), R3 = 1; otherwise, R3 = 0.

[0093] Step S323: Calculate the overall overlap R = α·R1 + β·R2 + γ·R3; where α, β, and γ are the weight coefficients of spatial region parameters, pollutant type parameters, and terrain adaptation parameters, respectively, and α + β + γ = 1. The weights are dynamically adjusted according to environmental characteristics. For complex terrain areas (such as mountains / canyons), the weight of α is increased to 0.5, and for areas with dense pollutants (such as industrial areas), the weight of β is increased to 0.5.

[0094] Step S324: If the overall overlap R is greater than or equal to the second preset threshold (e.g., 0.6), then the two subtasks are determined to meet the overlap condition and are merged.

[0095] The beneficial effects of adopting the above technical solutions are: (1) By “encoding data parsing - subtask overlap calculation - overlapping task merging - KDTree algorithm path generation”, the data transmission and path planning of the task are optimized, which solves the problem of path overlap and high redundancy in traditional UAV task planning. Overlap calculation and merging reduce invalid flights. KDTree algorithm adapts to the path generation requirements of two-dimensional scenes with complex terrain, improving the rationality and efficiency of path planning. It also reduces the energy consumption of UAV swarms, shortens the task execution time, and ensures that the monitoring area is covered without omissions. It overcomes the shortcomings of “insufficient precision of UAV task planning” in the existing technology and strengthens the practicality of collaborative monitoring. (2) Extract the three-dimensional core parameters of “spatial region + pollutant type + terrain adaptation”, calculate the comprehensive overlap by weighting, and dynamically adjust the weights to adapt to different environmental characteristics. This breaks through the traditional task merging logic based solely on spatial dimension, introduces terrain adaptation parameters to adapt to complex terrain differences, ensures the consistency of monitoring tasks by pollutant type parameters, and makes the merging judgment more accurate by weighting the comprehensive overlap. This avoids unreasonable merging of sub-tasks with different terrains and different pollution scenarios under complex terrain, improves the scientific nature of task allocation, reduces data redundancy and monitoring blind spots, and further enhances the accuracy of collaborative monitoring.

[0096] In the further optimized scheme of this embodiment, the specific execution process of the dynamic trajectory planning in step S4 is as follows:

[0097] Step S41: The main UAV acquires digital elevation model data of the planned flight path and target area, and extracts the coordinates and elevations of terrain feature points (such as peaks, valleys, and steep slopes).

[0098] Step S42: Construct a three-dimensional obstacle avoidance constraint model based on digital elevation model data, using the minimum safe flight altitude, turning radius, and remaining energy consumption of the sub-UAV as constraints.

[0099] Step S43: A heuristic search algorithm is used to introduce an energy consumption cost factor to perform three-dimensional correction on the planned flight path, generating a sub-drone execution route with "obstacle avoidance + optimal energy consumption"; where, the energy consumption cost factor = the current remaining power of the sub-drone / the estimated energy consumption to complete the route segment, and priority is given to allocating short-path, low-altitude and non-steep-slope execution routes to sub-drones with low power.

[0100] Step S44: If the sub-UAV detects a sudden terrain obstacle (such as a collapsed body or an unmarked steep slope) during flight, the main UAV receives the obstacle coordinates in real time, re-executes steps S41 to S43, generates a temporary alternative route, and sends it to the corresponding sub-UAV.

[0101] The beneficial effects of adopting the above technical solutions are as follows: a three-dimensional obstacle avoidance model is constructed based on DEM data, and an improved heuristic search algorithm is used to introduce an energy consumption cost factor to achieve dynamic trajectory planning that achieves "obstacle avoidance + optimal energy consumption". This supports temporary adjustments for sudden obstacles, accurately adapts to the characteristics of complex terrain such as peaks, valleys, and steep slopes, and solves the problems of weak obstacle avoidance capability and excessive energy consumption in traditional trajectory planning. The energy consumption factor prioritizes ensuring that low-battery electronic UAVs can complete their tasks, improving the endurance utilization rate of UAV swarms. The dynamic adjustment mechanism for sudden obstacles avoids mission interruption, ensuring the continuity and safety of monitoring tasks in complex terrain, and filling the technical gap in adapting existing UAV trajectory planning to complex terrain.

[0102] In the further optimized scheme of this embodiment, the process of constructing the pollution source tracing model in step S6 is as follows:

[0103] The flight locations and pollutant components (such as nitrogen oxides and nitrogen dioxide) of historical sub-tasks are used as inputs to the convolutional neural network, and the pollution source tracing results of historical sub-tasks are used as training labels.

[0104] By training a convolutional neural network with training labels, a pollution source tracing model can be obtained that can trace pollution sources in different environments.

[0105] The beneficial effects of adopting the above technical solution are as follows: Using "historical flight location + pollutant composition" as input and "historical source tracing results" as training labels, a pollution source tracing model is constructed through a convolutional neural network. This solves the problems of weak generalization ability and poor adaptability to different environments in traditional source tracing models. The model training data closely matches actual monitoring scenarios, improving adaptability to complex terrain and different pollution environments. Compared to single source tracing models, this model can fully explore the correlation characteristics between pollutants and pollution sources, providing reliable algorithmic support for subsequent quantitative calculations, enhancing the scientific rigor and accuracy of source tracing results, and addressing the pain point of "lack of data support in single source tracing models" in existing technologies.

[0106] In the further optimized scheme of this embodiment, the specific execution process of pollution source intensity calculation and source tracing in step S6 is as follows:

[0107] Step S61: The main UAV integrates the pollutant analysis conclusions, the flight location information of the sub-UAVs (such as latitude and longitude, altitude) and the basic parameters of the monitoring points (such as terrain slope, ambient temperature and humidity, wind speed), and generates a standardized source tracing calculation dataset after removing missing values ​​and outliers.

[0108] Step S62: Input the source tracing calculation dataset into the built-in pollution source tracing model based on convolutional neural network, and output the candidate pollution source list and preliminary weights for each pollutant at each monitoring point after feature extraction.

[0109] Step S63, according to formula C en =ω en ·S e +B n Calculate the contribution of candidate pollution sources to the corresponding pollutants; where C en Let ω be the contribution of the e-th pollution source to the n-th pollutant. en S represents the contribution ratio of the pollution source tracing model output based on pollutant transport patterns and complex terrain features. e B represents the pollutant emission intensity of the pollution source per unit time. n The background concentration of pollutants in the target area that is not subject to human pollution;

[0110] Step S64: For each pollutant at a single monitoring point, sort them from high to low according to their contribution value, screen out the top (e.g., the first) core pollution source, and combine the terrain features to clarify the transmission path of pollutants from the core pollution source to the monitoring point.

[0111] Step S65: Integrate the source tracing results of individual locations to generate a heat map of pollution source intensity distribution in the area covered by the drone and the overall source tracing conclusion.

[0112] The beneficial effects of adopting the above technical solution are: through "multi-dimensional data integration - model feature extraction - quantization formula calculation C"en- The "Core Pollution Source Screening - Heat Map Output" process refines the source tracing workflow, integrating topographic, location, and pollutant data to achieve comprehensive quantification and precision in the source tracing process. Multi-dimensional data integration eliminates the limitations of single data points in complex terrain. en The formula clearly defines the contribution of pollution sources, the screening of core pollution sources focuses on key pollution sources, and the heat map makes the source tracing results intuitive and easy to understand; it solves the problems of "vague results and lack of quantitative basis" in traditional source tracing, improves the accuracy of pollution source location and the clarity of transmission path, provides targeted solutions for pollution control, and the source tracing accuracy is significantly improved compared with traditional methods.

[0113] In the further optimized scheme of this embodiment, after step S6, a multi-machine cluster traceability data cross-validation step is also included, and the specific execution process is as follows:

[0114] Step S71: The ground control center collects the source tracing conclusions uploaded by all UAV groups and extracts the source tracing results of the same pollutant in the overlapping areas of the monitoring area from different UAV groups;

[0115] Step S72: Calculate the deviation rate of the pollution source contribution value of the same pollutant in the overlapping area. Deviation rate = |C calculated by Unit A en- Unit B's calculation of C en | / max(C calculated by unit A) en Unit B's calculation of C en ); where C en Let e ​​be the contribution of the e-th pollution source to the n-th pollutant;

[0116] Step S73: If the deviation rate is greater than the preset deviation threshold (e.g., 15%), then the high-resolution satellite remote sensing data and historical data from ground monitoring stations in the overlapping area are used as references, and the deviation source tracing results are calibrated by combining the Gaussian plume model corrected for complex terrain.

[0117] Step S74: Integrate the source tracing results of all calibrated units and regenerate the overall source tracing report for the target area.

[0118] The beneficial effects of adopting the above technical solutions are as follows: The newly added process of "multi-drone swarm cross-verification - deviation rate calculation - satellite / ground data calibration - terrain-corrected Gaussian plume model" verifies the source tracing results in overlapping areas, overcomes the limitations of single-drone swarm source tracing data, quantifies data consistency through deviation rate calculation, supplements monitoring blind spots under complex terrain through satellite and ground data calibration, and improves calibration accuracy through the terrain correction model. It also solves the problem of large data deviations between different drone swarms under complex terrain, resulting in a more consistent and accurate global source tracing report after integration, with source tracing accuracy improved by at least 20% compared to a single drone swarm, filling the technical gap in existing collaborative monitoring that "lacks cross-drone data verification."

[0119] In a further optimized version of this embodiment, the pollution source tracing model in step S6 is dynamically iteratively optimized during use. The specific process includes:

[0120] Step S661: The ground control center summarizes the actual detection data, source tracing results, and pollution source verification results of the UAV swarm on a weekly / monthly basis;

[0121] Step S662: Compare the verification results with the source tracing conclusions output by the model, and calculate the model source tracing accuracy rate; where, the model source tracing accuracy rate = number of core pollution sources confirmed by verification / number of core pollution sources output by the model;

[0122] Step S663: If the accuracy is less than the third preset threshold (e.g., 85%), then filter out samples with low accuracy (e.g., source data in canyon terrain or high humidity environment) and add them to the model training set.

[0123] Step S664: The original convolutional neural network model is fine-tuned using the transfer learning method (freezing the bottom feature extraction layer and updating only the top fully connected layer) to generate an iterative model adapted to the dynamic pollution features of complex terrain.

[0124] In step S665, the ground control center distributes the iterated model to all main UAVs, replacing the original pollution source tracing model and realizing the self-adaptive optimization of the model.

[0125] The beneficial effects of adopting the above technical solutions are: establishing a dynamic optimization mechanism of "regular data aggregation - accuracy evaluation - low-precision sample supplementation - transfer learning fine-tuning - model iteration and distribution", solving the adaptation decay problem caused by the "one-time training, lifelong use" of traditional source tracing models, calibrating the model through data verification by environmental protection departments, reducing retraining costs through transfer learning fine-tuning, and allowing the model to continuously adapt to the dynamic pollution characteristics of complex terrain through dynamic iteration; maintaining high source tracing accuracy of the model in the long term, avoiding the decline in source tracing accuracy due to changes in terrain and pollution type, and ensuring the long-term effectiveness and practicality of the technical solutions.

[0126] The following is a more detailed explanation of its capabilities.

[0127] This embodiment targets the area surrounding a mountainous canyon-type industrial park (target area: 103°25′-103°35′E, 29°40′-29°50′N, with an area of ​​approximately 50 km²). The terrain of this area is complex, including 3 main canyons and 5 low mountains with an altitude of 500-1200m. There are potential pollution sources such as chemical enterprises and coal-fired power plants, and it is necessary to achieve accurate detection and source tracing of multiple pollutants.

[0128] II. Equipment and Parameter Configuration

[0129] 1. Drone swarm configuration

[0130] Number of drone groups: 3 groups (each group contains 1 main drone + 4 sub-drones)

[0131] Main drone model: DJI M300RTK (40-minute flight time, maximum payload 2.7kg, supports DEM data import)

[0132] Sub-drone model: DJI Mavic 3 Enterprise (34-minute flight time, equipped with multispectral sensor + gas detection module)

[0133] Detection module parameters:

[0134] Gas sensor: Detects SO2 (range 0-50ppm, accuracy ±0.1ppm), NO X (0-100ppm, ±0.2ppm), VOC S (0-200ppm, ±0.5ppm), PM2.5 / PM10 (0-1000μg / m³, ±5μg / m³), NO2 (0-50ppm, ±0.1ppm), CO (0-1000ppm, ±1ppm)

[0135] Positioning accuracy: RTK centimeter-level positioning (latitude and longitude error ≤2cm, altitude error ≤5cm)

[0136] 2. Ground control center configuration

[0137] Hardware: Industrial-grade server (CPU: Intel Xeon Gold 6330, 64GB RAM, 2TB storage), multi-channel data receiving terminal

[0138] Software: Python 3.9 (with Scikit-learn and TensorFlow 2.8 frameworks), ArcGIS Pro (terrain analysis and path visualization), and a custom collaborative monitoring and management system.

[0139] 3. Core Parameter Presets

[0140] First preset threshold (difference in terrain complexity coefficient): 0.2

[0141] Second preset threshold (overall overlap): 0.6

[0142] Preset deviation threshold: 15%

[0143] Third preset threshold (model accuracy): 85%

[0144] Threshold for the contribution of core pollution sources: 10%

[0145] III. Specific Implementation Steps

[0146] Step S1: Layered and hierarchical task planning

[0147] The ground control center imports DEM data of the target area using ArcGIS Pro and plans the layered and height-based tasks:

[0148] Flight altitude stratification (by altitude):

[0149] Lower level: 0-100m (bottom of the canyon, low-altitude area around the industrial park)

[0150] Middle layer: 100-300m (middle of the hillside, the main area for pollutant diffusion)

[0151] High-rise buildings: 300-500m (mountain top, high-altitude pollutant transmission channel)

[0152] Monitoring task stratification:

[0153] Lower layer: Screening for nitrogen oxides (NOx) X The main pollutants are sulfur-containing compounds (SO2), with a sampling frequency of once every 10 seconds.

[0154] Middle layer: Precise detection of volatile organic compounds (VOCs) S The sample contains 12 components including benzene and toluene, and particulate matter (PM2.5 / PM10). The sampling frequency is once every 5 seconds.

[0155] High-rise buildings: High-sensitivity detection of nitrogen dioxide (NO2) and carbon monoxide (CO), sampling frequency 1 time / 3 seconds.

[0156] The ground control center sent the mission to three drone clusters (one cluster was responsible for the western canyon, two clusters were responsible for the central low mountains, and three clusters were responsible for the area around the eastern industrial park).

[0157] Step S2: Subtask Breakdown and Coding

[0158] Taking a group of main drones as an example, after receiving the hierarchical and altitude-based task, it is divided into 6 sub-tasks:

[0159] Subtask 1: Northern section of the western canyon, 0-100m, NO X +SO2 screening

[0160] Subtask 2: Southern section of the western canyon, 0-100m, NO X +SO2 screening

[0161] Subtask 3: 100-300m north section of the western canyon, VOC S PM2.5 / PM10 detection

[0162] Subtask 4: 100-300m south section of the western canyon, VOC SPM2.5 / PM10 detection

[0163] Subtask 5: NO2+CO detection in the northern section of the western canyon, 300-500m.

[0164] Subtask 6: NO2+CO detection in the southern section of the western canyon, 300-500m.

[0165] The main UAV encodes the six sub-tasks according to the agreed coding rules (such as "area number-altitude layer-contaminant type"), generating coded data (such as "WN-0-100-NO"). X +SO2”, sent to the ground control center.

[0166] Step S3: Task data transfer optimization and path generation

[0167] (1) Subtask analysis and overlap calculation

[0168] The ground control center receives coded data from three sets of main UAVs, parses it to obtain 18 sub-tasks, and extracts the three-dimensional core parameters of each sub-task:

[0169] Spatial region parameters: For example, the monitoring area boundary for sub-task 1 is 103°25′-103°28′ east longitude and 29°40′-29°45′ north latitude.

[0170] Pollutant type parameter: For example, the pollutant type set for subtask 1 is {NO} X SO2

[0171] Terrain adaptation parameters: Based on DEM data, the terrain complexity coefficient for subtask 1 (canyon bottom) is 0.8, and for subtask 5 (mountain top area) it is 0.3.

[0172] Calculate the overall overlap between any two subtasks, taking subtask 1 from group 1 and subtask 3 from group 2 (spatial partial overlap) as an example:

[0173] Spatial overlap degree R1 = overlap area (0.8 km²) / union area (5.2 km²) ≈ 0.15

[0174] Pollutant type overlap R² = Common pollutant types (0 types) / Total types (4 types) = 0

[0175] Terrain adaptation overlap R3 = |0.8 - 0.7| (The terrain complexity coefficient for subtask 3 in group 2 is 0.7) ≤ 0.2 → R3 = 1

[0176] The overall overlap ratio R = 0.5 × 0.15 + 0.5 × 0 + 0 × 1 = 0.075 < 0.6, therefore, no merging is allowed.

[0177] (2) Path generation

[0178] For the 18 unmerged subtasks, two-dimensional flight paths are generated in ArcGIS Pro using the KDTree algorithm. For example, the path for subtask 1 is "starting point (103°25′E, 29°40′N) → flying straight along the bottom of the canyon → passing 3 monitoring points → ending point (103°28′E, 29°45′N)", with a path length of 8.5km, avoiding known steep slope areas.

[0179] Step S4: Dynamic Track Planning and Data Acquisition

[0180] After receiving the planned flight path, the first group of master UAVs performs dynamic trajectory planning:

[0181] (1) Three-dimensional correction

[0182] The main UAV imports DEM data of the target area, extracts terrain feature points such as mountain peaks (elevation 1020m) and valley turning points (103°26′E, 29°42′N), and constructs a three-dimensional obstacle avoidance constraint model (minimum safe flight altitude 5m, turning radius 10m). An improved A* algorithm is used to introduce an energy consumption cost factor (e.g., if sub-UAV 1 has 80% remaining battery power and an estimated energy consumption of 30%, the energy consumption cost factor = 2.67), generating a three-dimensional corrected execution route: "Start point → bypass steep slope (elevation difference > 50m) → reduce flight altitude to 50m → pass monitoring points → endpoint", with a total length of 9.2km and an estimated energy consumption of 25%.

[0183] (2) Data collection

[0184] Four sub-drones flew according to their assigned sub-tasks. When sub-drone 1 performed sub-task 1, it collected data at three monitoring points:

[0185] Monitoring point 1 (103°26′E, 29°41′N): NO X Concentration 35.2 ppm, SO2 concentration 8.7 ppm

[0186] Monitoring point 2 (103°27′E, 29°43′N): NO X Concentration 38.5 ppm, SO2 concentration 9.3 ppm

[0187] Monitoring point 3 (103°28′E, 29°45′N): NO X Concentration 36.1 ppm, SO2 concentration 8.9 ppm

[0188] The sub-drone uploads data to the main drone in real time. The main drone performs preprocessing (noise reduction and format standardization) to remove the abnormal SO2 value at monitoring point 2 caused by airflow interference (the original data was 19.3 ppm, which exceeded the error range), thus forming complete pollution source information.

[0189] Step S5: Pollutant component analysis

[0190] The ground control center received pollution source information uploaded by three main UAVs, and performed component analysis using the gas chromatography-mass spectrometry (GC-MS) data analysis module to generate pollutant analysis conclusions:

[0191] Lower level: NO X The average concentration was 37.2 ppm (mainly NO and NO2, with NO2 accounting for approximately 35%), and the average SO2 concentration was 9.1 ppm.

[0192] Middle layer: VOC S The concentrations of benzene and toluene were 2.3 ppm, 1.8 ppm, 85 μg / m³ for PM2.5, and 120 μg / m³ for PM10.

[0193] High-rise buildings: Average NO2 concentration 12.5 ppm, average CO concentration 45 ppm

[0194] The ground control center sends the analysis results to the corresponding group's main UAV.

[0195] Step S6: Calculation and tracing of pollution source intensity

[0196] After receiving the pollutant analysis results, the first group of main UAVs activates the built-in pollution source tracing model (built based on CNN, with input dimensions of "flight position + pollutant composition + terrain parameters", and the training set contains 500 sets of historical source tracing data):

[0197] (1) Data integration and model input

[0198] Integrating pollutant analysis conclusions (NO X =37.2ppm, NO2=13.02ppm), the flight position of the sub-UAV (latitude and longitude + altitude) and the basic parameters of the monitoring point (terrain slope 35°, temperature 25℃, humidity 60%, wind speed 3m / s), generate a standardized dataset, and input it into the source tracing model.

[0199] (2) Calculation of contribution value

[0200] The model outputs a list of three candidate pollution sources and their preliminary weights, according to formula C. en =ω en ·S e +B n Calculate contribution value (B) n The default setting is NO. X =5ppm, NO2=2ppm):

[0201] Candidate pollution source 1 (a coal-fired power plant, E103°27′, N29°44′): ω en(NO) X =0.6, S e (NO) X =50ppm / h→C en =0.6×50+5=35ppm

[0202] Candidate pollution source 2 (a chemical plant, E103°26′, N29°43′): ω en (NO) X =0.3, S e (NO) X =40ppm / h→C en =0.3×40+5=17ppm

[0203] Candidate pollution source 3 (a car repair shop, E103°28′, N29°42′): ω en (NO) X =0.1, S e (NO) X =20ppm / h→C en =0.1×20+5=7ppm

[0204] (3) Results of screening and tracing of core pollution sources

[0205] Based on the contribution value ranked from highest to lowest, candidate pollution source 1 (contribution percentage 35 / (35+17+7)≈61%≥10%) and candidate pollution source 2 (contribution percentage 30%≥10%) were identified as the core pollution sources. Considering the terrain features (the canyon runs north-south, wind speed 3m / s), the pollutant transport paths were determined as follows: "Coal-fired power plant → diffuses northward along the canyon → monitoring points 1-3" and "Chemical enterprises → diffuses northeastward along the canyon → monitoring points 2-3".

[0206] The main UAV generates a heat map of pollution source intensity distribution in the area covered by the group (the coal-fired power plant area is a high-intensity pollution area with a concentration of ≥40ppm) and overall source tracing conclusions, and sends them to the ground control center.

[0207] Example 2

[0208] The present invention also provides a collaborative environmental monitoring and pollution source tracing system for unmanned aerial vehicle (UAV) swarms in complex terrain, to implement a collaborative environmental monitoring and pollution source tracing method for unmanned aerial vehicle (UAV) swarms in complex terrain as described in Example 1. The system is characterized by comprising: a ground control center, and at least one UAV swarm connected to the ground control center; each UAV swarm includes a master UAV and several sub-UAVs.

[0209] The ground control center is equipped with:

[0210] The task planning module is used to pre-plan the hierarchical and altitude-based tasks for the target area and send the hierarchical and altitude-based tasks to each corresponding drone cluster.

[0211] The encoding data processing module is used to receive the encoded data uploaded by the main UAV, optimize the data transmission of the encoded data, generate the planned flight path for each sub-task, and send the planned flight path to the corresponding group of main UAVs.

[0212] The pollutant analysis module is used to receive complete pollution source information uploaded by the main UAV, analyze the pollutant components, form pollutant analysis conclusions, and send the pollutant analysis conclusions to the corresponding group's main UAV;

[0213] The results management module is used to receive the pollution source tracing results uploaded by the main drone. After the drone returns, it summarizes and uploads the pollutant detection and tracing results to the server.

[0214] The main drone is equipped with:

[0215] The subtask processing module is used to receive hierarchical and high-level tasks, decompose them into multiple subtasks, encode the multiple subtasks into encoded data according to the agreed encoding rules, and send them to the ground control center.

[0216] The trajectory planning module is used to receive planned flight paths, combine hierarchical and altitude-based tasks to complete task allocation and dynamic trajectory planning, and control the sub-UAVs in the same group to perform data collection according to the assigned tasks and execution routes.

[0217] The data preprocessing module is used to receive pollution source-related data uploaded by the sub-UAV, preprocess it, form complete pollution source information, and then send it to the ground control center.

[0218] The source tracing calculation module, with a built-in pollution source tracing model, receives pollutant analysis results and, combined with flight location information, calculates the contribution value C of each pollutant source to each pollutant using a quantitative formula. en It also completes the calculation and tracing of pollution source intensity, generates pollution source tracing results, and sends them to the ground control center;

[0219] The sub-drone is equipped with:

[0220] The pollutant detection module is mounted on a sub-drone to collect pollution source-related data according to the assigned tasks and execution routes, and uploads the pollution source-related data to the main drone in real time.

[0221] The beneficial effects of adopting the above technical solution are as follows: The three-tiered modular architecture of "ground control center - main UAV - sub-UAV" ensures a one-to-one correspondence between functional modules and methodological steps, achieving coordinated linkage in task planning, data processing, and source tracing calculations. This solves the problems of "functional fragmentation and poor coordination" in existing monitoring systems. The modular design enhances system scalability and maintainability. The ground control center oversees the overall situation, the main UAV is responsible for local scheduling, and the sub-UAVs focus on data collection, resulting in clear hierarchy and division of labor. This ensures the full implementation of methodological requirements, improves the stability and operability of monitoring and source tracing in complex terrain, reduces deployment and operating costs compared to distributed systems, and enhances the efficiency of data transmission and processing, laying the foundation for large-scale application.

Claims

1. A method for collaborative environmental monitoring and pollution source tracing using unmanned aerial vehicle (UAV) swarms in complex terrain, characterized in that, Includes the following steps: Step S1: The ground control center pre-plans the hierarchical and height-based tasks for the target area and sends them to each corresponding UAV group; each UAV group contains one master UAV and the rest are sub-UAVs. In step S2, the main UAV of the corresponding group receives the hierarchical and altitude-divided task, decomposes it into multiple sub-tasks, and then encodes the multiple sub-tasks according to the agreed encoding rules to form encoded data, which is then sent to the ground control center. Step S3: The ground control center receives the coded data, optimizes the mission data transmission, generates the planned flight path for each sub-mission, and then sends the planned flight path to the master UAV of the corresponding group. In step S4, the main UAV of the corresponding group receives the planned flight path, combines the hierarchical and altitude-based tasks, performs task allocation and dynamic trajectory planning, so that the sub-UAVs in the same group collect pollution source-related data through the onboard pollutant detection module according to the assigned tasks and execution routes, and upload it to the main UAV in real time for preprocessing to form complete pollution source information, and send it to the ground control center. In step S5, the ground control center receives pollution source information, performs pollutant component analysis, forms pollutant analysis conclusions, and then sends them to the main UAV of the corresponding group. Step S6: The main UAV of the corresponding group receives the pollutant analysis results, combines them with the flight location information, and uses the built-in pollution source tracing model to calculate the contribution value C of each pollutant source to the pollutant source through a quantitative formula. en The system calculates and traces the intensity of pollution sources; after tracing the pollution sources, it sends the results to the ground control center; the drone returns, and the ground control center uploads the pollutant detection and tracing results to the server.

2. The method for collaborative environmental monitoring and pollution source tracing of UAV swarms in complex terrain according to claim 1, characterized in that, The layered and altitude-based tasks in step S1 include flight altitude layering and monitoring task layering: the flight altitude is divided into at least three layers from low to high, with different altitude layers corresponding to different monitoring tasks; among them, the low-level monitoring task is to screen the concentration of major pollutants including nitrogen oxides and sulfur-containing compounds, the mid-level monitoring task is to accurately detect the concentration of volatile organic compounds and particulate matter, and the high-level monitoring task is to detect nitrogen dioxide and carbon monoxide with high sensitivity.

3. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, The specific execution process in step S3 is as follows: Step S31: After receiving the coded data, the ground control center parses it according to the coding rules corresponding to the monitoring tasks under different environments to obtain multiple sub-tasks; Step S32: Calculate the overlap between any two subtasks, and merge subtasks whose overlap meets the preset conditions according to the task distribution under different environments. Step S33: Generate the planned flight paths of each subtask after merging in different scenarios on a two-dimensional plane using the KDTree algorithm, and then send the planned flight paths to the master UAV of the corresponding group.

4. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 3, characterized in that, The specific execution process in step S32 is as follows: Step S321: Extract the core feature parameters of each subtask. The core feature parameters include spatial region parameters representing the boundary coordinates of the monitoring area corresponding to the subtask, pollutant type parameters representing the set of pollutant types to be monitored by the subtask, and terrain adaptation parameters representing the terrain complexity coefficient of the subtask area. Step S322: Based on spatial region parameters, pollutant type parameters, and terrain adaptation parameters, calculate the spatial overlap degree R1, pollutant type overlap degree R2, and terrain adaptation overlap degree R3, respectively; where, Spatial overlap R1 = Intersection area of ​​the two sub-task monitoring areas / Union area of ​​the two sub-task monitoring areas; Pollutant type overlap R2 = Number of pollutant types jointly monitored by the two sub-tasks / Total number of pollutant types monitored by the two sub-tasks after deduplication; Terrain adaptation overlap R3: When the difference in terrain complexity coefficient between two subtasks is less than or equal to the first preset threshold, R3 = 1; otherwise, R3 = 0. Step S323: Calculate the overall overlap R = α·R1 + β·R2 + γ·R3; where α, β, and γ are the weight coefficients of spatial region parameters, pollutant type parameters, and terrain adaptation parameters, respectively, and α + β + γ = 1. The weights are dynamically adjusted according to environmental characteristics. For complex terrain areas, the weight of α is increased to 0.5, and for pollutant-dense areas, the weight of β is increased to 0.

5. Step S324: If the overall overlap R ≥ the second preset threshold, then the two subtasks are determined to meet the overlap condition and are merged.

5. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, The specific execution process of the trajectory dynamic planning in step S4 is as follows: Step S41: The main UAV acquires digital elevation model data of the planned flight path and target area, and extracts the coordinates and elevations of terrain feature points; Step S42: Construct a three-dimensional obstacle avoidance constraint model based on digital elevation model data, using the minimum safe flight altitude, turning radius, and remaining energy consumption of the sub-UAV as constraints. Step S43: A heuristic search algorithm is used to introduce an energy consumption cost factor to perform three-dimensional correction on the planned flight path, generating a sub-drone execution route with "obstacle avoidance + optimal energy consumption"; where, the energy consumption cost factor = the sub-drone's current remaining power / the estimated energy consumption to complete the route segment, and priority is given to allocating short-path, low-altitude and non-steep-slope execution routes to sub-drones with low power. Step S44: If the sub-UAV detects a sudden terrain obstacle during flight, the main UAV receives the obstacle coordinates in real time, re-executes steps S41 to S43, generates a temporary alternative route, and sends it to the corresponding sub-UAV.

6. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, The process of constructing the pollution source tracing model in step S6 is as follows: The flight locations and pollutant components of historical sub-tasks are used as inputs to the convolutional neural network, and the pollution source tracing results of historical sub-tasks are used as training labels. By training a convolutional neural network with training labels, a pollution source tracing model can be obtained that can trace pollution sources in different environments.

7. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, The specific execution process for calculating and tracing the pollution source intensity in step S6 is as follows: Step S61: The main UAV integrates the pollutant analysis conclusions, the flight position information of the sub-UAVs and the basic parameters of the monitoring points, and generates a standardized source tracing calculation dataset after removing missing values ​​and outliers. Step S62: Input the source tracing calculation dataset into the built-in pollution source tracing model based on convolutional neural network, and output the candidate pollution source list and preliminary weights for each pollutant at each monitoring point after feature extraction. Step S63, according to formula C en =ω en ·S e +B n Calculate the contribution of candidate pollution sources to the corresponding pollutants; where C en Let ω be the contribution of the e-th pollution source to the n-th pollutant. en S represents the contribution ratio of the pollution source tracing model output based on pollutant transport patterns and complex terrain features. e B represents the pollutant emission intensity of the pollution source per unit time. n The background concentration of pollutants in the target area that is not subject to human pollution; Step S64: For each pollutant at a single monitoring point, sort them from high to low contribution value, screen out the top core pollution sources, and determine the transmission path of pollutants from the core pollution sources to the monitoring points in combination with topographic features. Step S65: Integrate the source tracing results of individual locations to generate a heat map of pollution source intensity distribution in the area covered by the drone and the overall source tracing conclusion.

8. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, Following step S6, a multi-machine cluster traceability data cross-validation step is also included, the specific execution process of which is as follows: Step S71: The ground control center collects the source tracing conclusions uploaded by all UAV groups and extracts the source tracing results of the same pollutant in the overlapping areas of the monitoring area from different UAV groups; Step S72: Calculate the deviation rate of the pollution source contribution value of the same pollutant in the overlapping area. Deviation rate = |C calculated by Unit A en- Unit B's calculation of C en | / max(C calculated by unit A) en Unit B's calculation of C en ); where C en Let e ​​be the contribution of the e-th pollution source to the n-th pollutant; Step S73: If the deviation rate is greater than the preset deviation threshold, the high-resolution satellite remote sensing data and historical data of the ground monitoring station in the overlapping area are used as references, and the deviation source tracing results are calibrated by combining the Gaussian plume model corrected for complex terrain. Step S74: Integrate the source tracing results of all calibrated units and regenerate the overall source tracing report for the target area.

9. The method for collaborative environmental monitoring and pollution source tracing of unmanned aerial vehicle (UAV) swarms in complex terrain according to claim 1, characterized in that, In step S6, the pollution source tracing model undergoes dynamic iterative optimization during use. The specific process includes: Step S661: The ground control center summarizes the actual detection data, source tracing results, and pollution source verification results of the UAV swarm on a weekly / monthly basis; Step S662: Compare the verification results with the source tracing conclusions output by the model, and calculate the model source tracing accuracy rate; where, the model source tracing accuracy rate = number of core pollution sources confirmed by verification / number of core pollution sources output by the model; Step S663: If the accuracy is less than the third preset threshold, then the samples with low accuracy are selected and added to the model training set. Step S664: The original convolutional neural network model is fine-tuned using the transfer learning method to generate an iterative model adapted to the dynamic pollution characteristics of complex terrain. In step S665, the ground control center distributes the iterated model to all main UAVs, replacing the original pollution source tracing model and realizing the self-adaptive optimization of the model.

10. A collaborative environmental monitoring and pollution source tracing system for unmanned aerial vehicle (UAV) swarms in complex terrain, for implementing the collaborative environmental monitoring and pollution source tracing method for unmanned aerial vehicle (UAV) swarms in complex terrain as described in any one of claims 1 to 9, characterized in that, include: A ground control center, and at least one drone swarm connected to the ground control center; each drone swarm consists of a master drone and several sub-drones; The ground control center is equipped with: The task planning module is used to pre-plan the hierarchical and altitude-based tasks for the target area and send the hierarchical and altitude-based tasks to each corresponding drone cluster. The encoding data processing module is used to receive the encoded data uploaded by the main UAV, optimize the data transmission of the encoded data, generate the planned flight path for each sub-task, and send the planned flight path to the corresponding group of main UAVs. The pollutant analysis module is used to receive complete pollution source information uploaded by the main UAV, analyze the pollutant components, form pollutant analysis conclusions, and send the pollutant analysis conclusions to the corresponding group's main UAV; The results management module is used to receive the pollution source tracing results uploaded by the main drone. After the drone returns, it summarizes and uploads the pollutant detection and tracing results to the server. The main drone is equipped with: The subtask processing module is used to receive hierarchical and high-level tasks, decompose them into multiple subtasks, encode the multiple subtasks into encoded data according to the agreed encoding rules, and send them to the ground control center. The trajectory planning module is used to receive planned flight paths, combine hierarchical and altitude-based tasks to complete task allocation and dynamic trajectory planning, and control the sub-UAVs in the same group to perform data collection according to the assigned tasks and execution routes. The data preprocessing module is used to receive pollution source-related data uploaded by the sub-UAV, preprocess it, form complete pollution source information, and then send it to the ground control center. The source tracing calculation module, with a built-in pollution source tracing model, receives pollutant analysis results and, combined with flight location information, calculates the contribution value C of each pollutant source to the pollutant using a quantitative formula. en It also completes the calculation and tracing of pollution source intensity, generates pollution source tracing results, and sends them to the ground control center; The sub-drone is equipped with: The pollutant detection module is used to carry out the sub-drone to collect pollution source-related data according to the assigned tasks and execution routes, and upload the pollution source-related data to the main drone in real time.

Images