Three-dimensional pollutant monitoring method and system combined with wind field perception

Abstract

The invention provides a pollutant three-dimensional monitoring method and system combined with wind field perception, and relates to the technical field of pollutant monitoring, and the method comprises the steps: enabling obtained three-dimensional wind field scene data to be aligned with digital elevation model data, and obtaining fusion grid data of a monitoring region; determining a sampling point layout strategy according to the fused grid data, a preset wind field weight coefficient and a preset terrain weight coefficient; taking the three-dimensional wind field scene data and the acquired sampling data as input of a pre-constructed matching model to obtain an interpolation strategy of the monitoring area; based on the three-dimensional wind field scene data, the digital elevation model data and the sampling point layout strategy, performing optimization planning on an unmanned aerial vehicle sampling path by adopting a multi-target optimization algorithm to obtain a target three-dimensional route; and performing spatial interpolation operation on the target pollutant concentration data by adopting an interpolation strategy, and reconstructing a three-dimensional pollutant concentration distribution model of the monitoring area. According to the method provided by the invention, three-dimensional monitoring of pollutants in the monitoring area is realized.

Description

Technical Field

[0001] This application relates to the field of pollutant monitoring technology, and in particular to a three-dimensional pollutant monitoring method and system that combines wind field sensing. Background Technology

[0002] Air pollution is one of the major environmental problems facing the world, with profound impacts on human health and ecosystems. Effective control and reduction of air pollution are essential. Air pollutants mainly originate from human activities such as industrial production, vehicle exhaust, and coal combustion, as well as natural phenomena such as volcanic eruptions and forest fires. These pollutants include particulate matter, sulfur dioxide, nitrogen oxides, and volatile organic compounds.

[0003] Currently, traditional fixed monitoring stations are often sparsely and unevenly distributed due to high construction costs and complex maintenance, making it difficult to fully capture the dynamic diffusion characteristics of pollutants in three-dimensional space. Especially under complex urban wind fields or special terrain conditions in industrial areas, the actual transport paths and concentration distribution of pollutants often deviate significantly from the monitoring data. Summary of the Invention

[0004] In view of this, this application proposes a three-dimensional monitoring method and system for pollutants that combines wind field sensing.

[0005] Firstly, this application provides a three-dimensional pollutant monitoring method that combines wind field sensing, including: Acquire three-dimensional wind field scene data and digital elevation model data of the monitoring area, and align the three-dimensional wind field scene data to the same spatial resolution as the digital elevation model data to obtain fused grid data of the monitoring area. The three-dimensional wind field scene data includes wind field structure sub-region information divided from the monitoring area. The sampling point layout strategy is determined based on the fused grid data, the preset wind field weight coefficient, and the preset terrain weight coefficient. The three-dimensional wind field scene data and the acquired sampling data are used as inputs to a pre-built matching model to obtain the interpolation strategy for the monitoring area. Based on 3D wind field scene data, digital elevation model data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the sampling path of the UAV, and obtain the target 3D flight path. The interpolation strategy is used to perform spatial interpolation on the target pollutant concentration data to reconstruct a three-dimensional pollutant concentration distribution model of the monitoring area. The target pollutant concentration data is the sampling data obtained by the UAV flying along the three-dimensional flight path of the target.

[0006] In one embodiment, acquiring the three-dimensional wind field scene data of the monitoring area includes: Acquire the collected multidimensional meteorological data; The multidimensional meteorological data is preprocessed to obtain wind field optimization data, wherein the preprocessing includes smoothing filtering and outlier removal; The optimized wind field data is input into a pre-constructed wind field structure recognition model to classify regional wind field features, thereby obtaining three-dimensional wind field scene data of the monitored area.

[0007] In one embodiment, determining the sampling point layout strategy based on the fused grid data, a preset wind field weighting coefficient, and a preset terrain weighting coefficient includes: Based on the fused grid data, preset wind field weighting coefficient, and preset terrain weighting coefficient, the sampling priority value of each grid point in the monitoring area is calculated to obtain a sampling priority matrix. The sampling point distribution set of each wind field structure sub-region is determined based on the fused grid data, the sampling priority matrix, and the sampling optimization strategy corresponding to each wind field structure label. The sampling point layout strategy for the monitoring area is determined by combining the sampling point distribution set of each wind field structure sub-region.

[0008] In one embodiment, the wind field structure label corresponding to the wind field structure sub-region includes one of the following: downwind diffusion region, shear variation region, and static stagnation region; The interpolation algorithm corresponding to the downwind diffusion zone is wind vector weighted kriging interpolation, wherein the wind vector weighted kriging interpolation algorithm is a kriging interpolation algorithm that introduces wind speed and wind direction as covariates into a semi-variogram model. The interpolation algorithm for the shear variation zone is an adaptive kernel density estimation algorithm, and the kernel function bandwidth of the adaptive kernel density estimation algorithm is inversely correlated with the degree of regional concentration change. The interpolation algorithm for the static and stagnant region is an improved thin-plate spline interpolation algorithm, which is a thin-plate spline interpolation algorithm that adds a terrain gradient penalty term to the spline interpolation.

[0009] In one embodiment, the step of optimizing the UAV sampling path using a multi-objective optimization algorithm based on three-dimensional wind field scene data, digital elevation model data, and sampling point layout strategy to obtain the target three-dimensional flight path includes: With spatial coverage, wind field characteristic error, and range energy consumption as optimization objectives, a non-dominated sorting genetic algorithm is used to optimize and plan the sampling path of the UAV, thereby obtaining a target three-dimensional route that satisfies the sampling point layout strategy.

[0010] In one embodiment, the method further includes: K-fold cross-validation is used to evaluate the accuracy of the interpolation results and to calculate the error index within the monitoring area. The error index includes at least one of the root mean square error, mean absolute error, and spatial consistency index. A spatial distribution heatmap of errors is generated based on the error indicators within the monitoring area; The optimization area is determined based on the error spatial distribution heatmap, and the target optimization strategy is determined based on the error level of the optimization area. The optimization strategy includes at least one of parameter-level optimization, structure-level optimization, and sampling-level optimization. The parameter-level optimization is to adjust the parameter configuration of the current interpolation model according to preset rules. The structure-level optimization is to switch the interpolation model according to the error distribution characteristics. The sampling-level optimization is to plan additional UAV supplementary sampling routes for high error areas.

[0011] In one embodiment, the wind field structure label includes a downwind diffusion zone, a shear variation zone, and a static stagnation zone; the sampling optimization strategy corresponding to each wind field structure label includes: increasing the density of sampling points along the main wind direction axis in the downwind diffusion zone with a first density increase value; increasing the sampling coverage in the critical region of the wind speed gradient in the shear variation zone with a second density increase value; maintaining a preset sampling density in the static stagnation zone; and setting sampling points in layers along the vertical direction in areas with severe topographic undulations to form a three-dimensional sampling network with multi-height layers.

[0012] Secondly, this application also provides a three-dimensional pollutant monitoring system that combines wind field perception, including: a drone and a processing module; The processing module is communicatively connected to the UAV and is used to control the flight trajectory of the UAV and to execute the three-dimensional pollutant monitoring method combined with wind field perception as described in the first aspect.

[0013] In one embodiment, the processing module includes: a meteorological sensing module, a scheduling module, a processing module, and a communication module; The meteorological sensing module is used to collect meteorological data; The processing module is connected to the meteorological sensing module, and the processing module is used to execute the three-dimensional pollutant monitoring method combined with wind field sensing as described in the first aspect. The scheduling module is used to receive the target three-dimensional flight path output by the processing module, and coordinate multiple UAVs to work together based on the target three-dimensional flight path, so that the flight path of each UAV covers the target three-dimensional flight path. The communication module is used to enable communication between the scheduling module and each UAV.

[0014] In one embodiment, the scheduling module is used to divide the monitoring area into multiple sub-sampling areas according to altitude layers, generate sub-flight routes corresponding to each sub-sampling area, and control each UAV to cruise and sample along the corresponding sub-flight routes of each sub-sampling area.

[0015] The three-dimensional pollutant monitoring method combining wind field sensing proposed in this application has the following advantages over related technologies: 1. The three-dimensional pollutant monitoring method combined with wind field perception in this application obtains three-dimensional wind field scene data and digital elevation model data of the monitoring area, and aligns the three-dimensional wind field scene data to the same spatial resolution as the digital elevation model data to obtain fused grid data of the monitoring area, thereby realizing the alignment and fusion of wind field features and terrain features. Based on this, the sampling point layout strategy is first determined according to the fused grid data, preset wind field weight coefficients, and preset terrain weight coefficients. Then, the 3D wind field scene data and the acquired sampling data are used as inputs to a pre-built matching model to obtain the interpolation strategy for the monitoring area. Based on the 3D wind field scene data, digital elevation model data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the UAV sampling path to obtain the target 3D flight path. The target 3D flight path obtained in this way is comprehensive, efficient, and safe, providing a strong guarantee for the UAV to complete the monitoring task with high quality. After determining the target 3D flight path, the UAV can be controlled to sample along the target 3D flight path to obtain high-quality target pollutant concentration data. Finally, the interpolation strategy is used to perform spatial interpolation calculations on the target pollutant concentration data to reconstruct the 3D pollutant concentration distribution model of the monitoring area, thereby comprehensively capturing the dynamic diffusion characteristics of pollutants in 3D space and realizing 3D monitoring of pollutants in the monitoring area. In addition, since the pollutant 3D monitoring method of this application combines wind field characteristics and terrain characteristics, it can be widely applied to complex scenarios such as urban street valleys, industrial parks, and windy port areas.

[0016] 2. By increasing the sampling density along the prevailing wind axis in the downwind diffusion zone, rapid changes in pollutant concentrations along the windward diffusion direction can be captured. In shear variation zones with strong shear and turbulence, increasing the number of sampling points in the surrounding and downstream areas can ensure sufficient monitoring coverage for potential local vortices and concentration anomalies. For stagnant zones with low wind speeds and easy pollutant accumulation, deploying necessary monitoring points to observe pollutant retention and accumulation characteristics can prevent missing high-concentration accumulation phenomena due to prolonged calm winds. Furthermore, in areas with significant topographical differences and dramatic undulations, evenly distributing sampling points at different altitudes can reflect the impact of altitude changes on pollutant concentrations. Thus, while ensuring overall spatial coverage, this sampling scheme highlights key wind-driven areas and takes into account topographical factors, providing a scientific basis for subsequent UAV flight path planning. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying 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.

[0018] Figure 1 This is a flowchart illustrating a three-dimensional pollutant monitoring method incorporating wind field sensing in one embodiment of this application. Figure 2 This is a schematic diagram illustrating the fusion of wind field sensing data in one embodiment of this application; Figure 3 This is a flowchart illustrating step S102 in one embodiment of this application; Figure 4 This is a schematic diagram of the structure of a three-dimensional pollutant monitoring system incorporating wind field sensing in one embodiment of this application; Figure 5 This is a schematic diagram of a three-dimensional pollutant monitoring system that incorporates wind field sensing in another embodiment of this application. Detailed Implementation

[0019] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] In some embodiments, such as Figure 1 As shown, this application provides a three-dimensional pollutant monitoring method that combines wind field perception, which includes the following steps S101 to S105.

[0021] S101: Acquire 3D wind field scene data and digital elevation model data of the monitoring area, and align the 3D wind field scene data to the same spatial resolution as the digital elevation model data to obtain fused grid data of the monitoring area. The 3D wind field scene data includes information on the wind field structure sub-regions divided within the monitoring area.

[0022] In applications, multiple meteorological data can be monitored using meteorological sensors (such as miniature meteorological sensors deployed in drone nests, wind speed and direction measurement units, and auxiliary ground meteorological stations). Alternatively, meteorological data for the monitored area can be directly acquired via network connection. Based on this meteorological data, the wind field structure of the monitored area can be classified and labeled, resulting in three-dimensional wind field scene data for the monitored area. Wind field labels can include labels for downwind diffusion zones, shear variation zones, and static stagnant zones. The digital elevation model data for the monitored area can be determined based on the corresponding satellite remote sensing data.

[0023] After acquiring the 3D wind field scene data and digital elevation model (DEM) data of the monitoring area, the 3D wind field scene data can be interpolated or resampled to align with the spatial grid resolution of the DEM, thus constructing a unified analysis grid. For example, if the grid resolution of the DEM is 10 m × 10 m, the wind field data can be interpolated at the corresponding scale to ensure that information such as wind speed and direction corresponds one-to-one with the terrain elevation grid. After this processing, a grid dataset covering the monitoring area can be obtained, namely, the fused grid data of the monitoring area. Each grid node simultaneously possesses wind field characteristic values ​​(such as wind speed gradient and wind direction) and terrain characteristic values ​​(such as elevation difference and slope).

[0024] S102: Determine the sampling point layout strategy based on the fused grid data, preset wind field weight coefficient, and preset terrain weight coefficient.

[0025] Among them, wind speed gradient and topographic relief change rate can be selected as key indicators to reflect the importance of monitoring, and preset wind field weight coefficient and preset topographic weight coefficient can be determined based on these.

[0026] By multiplying the wind field feature value and terrain feature value of each grid in the fused grid by their respective weighting coefficients and summing them, the sampling priority of each grid unit is obtained. Finally, sampling resources are allocated according to priority, which allows sampling points to be densely distributed in key wind field areas and complex terrain areas, and sparsely distributed in areas with stable wind fields and gentle terrain. This ensures both accurate capture of core monitoring elements and sampling efficiency.

[0027] S103: Using the three-dimensional wind field scene data and the acquired sampling data as input to the pre-built matching model, the interpolation strategy for the monitoring area is obtained.

[0028] The matching model can be a decision model built based on decision logic (such as a rule tree) or a lightweight machine learning model.

[0029] In the application, multiple interpolation algorithms can be pre-set, and then a matching model can be pre-built. The input of the matching model is three-dimensional wind field scene data and the acquired sampling data. The output is the type of interpolation algorithm suitable for the wind field structure sub-region and its initial parameter configuration, thereby realizing the dynamic adaptive selection of the interpolation method.

[0030] S104: Based on 3D wind field scene data, digital elevation model data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the UAV sampling path to obtain the target 3D flight path.

[0031] It is understandable that 3D wind field data is used to quantify the impact of wind fields on UAV energy consumption and to impose safety constraints on UAVs. Digital elevation model (DEM) data defines the lower limit of flight altitude and obstacle avoidance range through terrain elevation information; the sampling point layout strategy clarifies the sampling points that must be covered. Based on this, objectives such as UAV energy consumption, path length, and sampling point coverage can be optimized. First, the 3D wind field scene data and DEM data can be converted into a constraint matrix, and the sampling point layout can be converted into a set of must-reach points. By determining the initial path population, the objective function value of each path is calculated; then, iterative optimization is performed through crossover and mutation operations, retaining the Pareto optimal solution; finally, based on actual needs (such as prioritizing energy saving or safety), a target 3D flight path is selected from the optimal solution. This flight path can complete data collection according to the sampling point layout and adapt to wind field and terrain conditions, achieving efficient and safe UAV sampling operations.

[0032] S105: Spatial interpolation calculations are performed on the target pollutant concentration data using an interpolation strategy to reconstruct a three-dimensional pollutant concentration distribution model of the monitoring area. The target pollutant concentration data is the sampling data obtained by the UAV flying along the target three-dimensional flight path.

[0033] Based on the interpolation strategy for the monitoring area determined in step S103, spatial interpolation calculations are performed on the pollutant concentration monitoring data (i.e., target pollutant concentration data) collected along the planned path in step S104. This allows for the prediction of pollutant concentration values ​​in unsampled areas, thereby generating a three-dimensional pollutant concentration distribution map of the monitoring area. This interpolation modeling process uses measured data from sampling points as input and infers the concentration distribution of the entire area by analyzing the spatial correlation between measurement points, achieving the goal of predicting pollutant concentrations in unknown areas based on limited sampling data.

[0034] The aforementioned three-dimensional pollutant monitoring method combining wind field perception acquires three-dimensional wind field scene data and digital elevation model (DEM) data of the monitoring area. The three-dimensional wind field scene data is aligned to the same spatial resolution as the DEM data to obtain fused grid data of the monitoring area, thus achieving alignment and fusion of wind field features and terrain features. Based on this, a sampling point layout strategy is first determined according to the fused grid data, preset wind field weight coefficients, and preset terrain weight coefficients. Then, the three-dimensional wind field scene data and the acquired sampling data are used as inputs to a pre-constructed matching model to obtain an interpolation strategy for the monitoring area. Based on the three-dimensional wind field scene data, DEM data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the UAV sampling path, obtaining a target three-dimensional flight path. This method provides comprehensive, efficient, and safe coverage of the target three-dimensional flight path, providing strong support for the UAV to complete the monitoring task with high quality. After determining the target three-dimensional flight path, the UAV can be controlled to sample along the target three-dimensional flight path to obtain high-quality target pollutant concentration data. Finally, an interpolation strategy is used to perform spatial interpolation calculations on the target pollutant concentration data to reconstruct a three-dimensional pollutant concentration distribution model of the monitoring area, thereby comprehensively capturing the dynamic diffusion characteristics of pollutants in three-dimensional space and realizing three-dimensional monitoring of pollutants in the monitoring area. In addition, since the pollutant three-dimensional monitoring method of this application combines wind field characteristics and terrain characteristics, it can be widely applied to complex scenarios such as urban street valleys, industrial parks, and windy port areas.

[0035] In some embodiments, step S101, acquiring three-dimensional wind field scene data of the monitoring area, includes: acquiring collected multidimensional meteorological data; preprocessing the multidimensional meteorological data to obtain optimized wind field data; and inputting the optimized wind field data into a pre-constructed wind field structure recognition model for regional wind field feature classification to obtain three-dimensional wind field scene data of the monitoring area. The preprocessing includes smoothing filtering and outlier removal.

[0036] In this embodiment, a wind field sensing module deployed in the drone nest can collect multi-dimensional meteorological data in real time, including wind speed, wind direction, and turbulence intensity. The wind field sensing module can use ultrasonic wind measurement technology for data acquisition. Figure 2 As shown, wind field sensing data can also be obtained by fusing multi-source data. Figure 2 This refers to the assimilation process of data from ground stations, drones, and numerical models.

[0037] For example, the smoothing filter uses a moving weighted average algorithm with a preset time window (e.g., 60 seconds) and weight coefficients distributed exponentially, effectively reducing drastic fluctuations in wind field data over short timescales and improving data stability. Outlier removal is based on the 3σ criterion, performing outlier analysis on the current observation data to remove outliers that deviate from the historical mean by more than a set tolerance, thus avoiding misleading subsequent sampling and optimization modules. Preprocessing the collected raw wind field data can improve the effectiveness and stability of the data.

[0038] Then, the preprocessed wind field data is input into a wind field structure identification model for regional wind field feature classification. This model comprehensively utilizes methods such as wind rose diagram statistical analysis and turbulence intensity threshold discrimination to evaluate the prevailing wind direction characteristics and turbulence level within the monitoring area, thereby dividing different types of wind field structure sub-regions. For example, the model can identify typical regional types such as the downwind diffusion zone downstream of the prevailing wind direction (the downwind diffusion channel formed by pollutants along the prevailing wind direction), the shear variation zone where wind speed changes rapidly with space and is prone to generating eddies, and the stagnant zone where low near-surface wind speed and stable air lead to easy accumulation of pollutants. The model assigns corresponding wind field structure labels to each sub-region and outputs high-resolution three-dimensional wind field scene data corresponding to the actual geographic space.

[0039] In some embodiments, such as Figure 3 As shown, in step S102, the sampling point layout strategy is determined based on the fused grid data, the preset wind field weight coefficient, and the preset terrain weight coefficient, including the following steps S301 to S303.

[0040] S301: Based on the fused grid data, preset wind field weight coefficient and preset terrain weight coefficient, the sampling priority value of each grid point in the monitoring area is calculated to obtain the sampling priority matrix.

[0041] The priority matrix is ​​essentially a two-dimensional matrix corresponding to the regional grid, and its size depends on the number of grids after discretization of the monitoring area. For example, wind speed gradient and topographic relief rate can be selected as key indicators reflecting the importance of monitoring, and assigned weights of 0.6 and 0.4 respectively (the weight values ​​can be adjusted according to the specific application scenario). For each grid point, its sampling priority value is calculated as 0.6 × (normalized wind speed gradient) + 0.4 × (normalized topographic relief rate). After the above calculation, the higher the value in the matrix, the more drastic the wind field changes and the higher the topographic complexity, i.e., the more critical the location is for the monitoring task, and sampling points should be placed there first.

[0042] S302: Determine the set of sampling points for each wind field structure sub-region based on the fused grid data, the sampling priority matrix, and the sampling optimization strategy corresponding to each wind field structure label.

[0043] S303: Determine the sampling point layout strategy for the monitoring area by combining the sampling point distribution set of each wind field structure sub-region.

[0044] After determining the priority matrix, the sampling point layout can be dynamically optimized based on the fused grid data, the sampling priority matrix, and the labels of each wind field structure. According to the calculated sampling priority matrix and the sampling optimization strategy corresponding to the wind field structure labels, the spatial distribution scheme of the sampling points can be dynamically adjusted to determine the sampling point distribution set of each wind field structure sub-region; combined with the sampling point distribution set of each wind field structure sub-region, the sampling point layout strategy of the monitoring area is determined. Thus, while ensuring overall spatial coverage, this scheme highlights key wind-driven areas and takes into account terrain factors, providing a scientific basis for subsequent UAV flight path planning.

[0045] In some embodiments, the wind field structure label includes a downwind diffusion zone, a shear variation zone, and a static stagnation zone; the sampling optimization strategy corresponding to each wind field structure label includes: increasing the density of sampling points along the main wind direction axis in the downwind diffusion zone by a first density increase value; increasing the sampling coverage in the critical region of the wind speed gradient in the shear variation zone by a second density increase value; maintaining a preset sampling density in the static stagnation zone; and setting sampling points in layers along the vertical direction in areas with severe topographic undulations to form a three-dimensional sampling network with multi-height layers.

[0046] It is understandable that increasing the sampling density along the prevailing wind axis in downwind diffusion areas such as the main leeward diffusion channels can capture rapid changes in pollutant concentrations along the windward diffusion direction. In shear fluctuation zones with strong shear and turbulence, increasing the number of sampling points in their surrounding and downstream areas ensures sufficient monitoring coverage for potential local eddies and concentration anomalies. For stagnant areas with low wind speeds and easy pollutant accumulation, necessary monitoring points are deployed to observe the retention and accumulation characteristics of pollutants, preventing the loss of high-concentration accumulation phenomena due to prolonged calm winds. In addition, for areas with significant topographic differences and dramatic undulations, evenly distributing sampling points at different height levels—such as valleys and building-street valleys—to form a multi-layered vertical sampling network ("upper-middle-lower") can reflect the impact of altitude changes on pollutant concentrations.

[0047] In some embodiments, the wind field structure label corresponding to the wind field structure sub-region includes one of the following: downwind diffusion region, shear variation region, and static stagnation region.

[0048] The interpolation algorithm corresponding to the downwind diffusion zone is wind vector weighted kriging interpolation, which is a kriging interpolation algorithm that introduces wind speed and wind direction as covariates into the semi-variogram model.

[0049] When Kriging interpolation is used, the system introduces a semivariogram model with wind vector direction constraints. Wind speed and wind direction are embedded as covariates in the modeling process of the semivariogram, so that the semivariogram value varies with wind direction, making the interpolation results more sensitive to wind direction. Under extreme calm wind conditions, the semivariogram will automatically degenerate into an isotropic form to ensure that the model does not overfit.

[0050] For example, the semi-variogram model of wind vector weighted Kriging interpolation is as follows:

[0051] in The adjustment coefficient is dynamically adjusted based on wind speed. ( (Wind speed, unit: m / s).

[0052] The interpolation algorithm for the shear variation zone is an adaptive kernel density estimation algorithm. The kernel function bandwidth of the adaptive kernel density estimation algorithm is inversely correlated with the degree of regional concentration change.

[0053] In application, by analyzing the dispersion of pollutant sampling values ​​and the intensity of local turbulence, the bandwidth parameter of the kernel function is dynamically set. A smaller bandwidth is used in areas where the observed values ​​change drastically and the local concentration peaks are significant, in order to improve spatial resolution; while a larger bandwidth is used in areas where the concentration distribution is relatively uniform, in order to smooth out random errors.

[0054] The interpolation algorithm for the static and stagnant region is an improved thin-plate spline interpolation algorithm, which is a thin-plate spline interpolation algorithm that adds a terrain gradient penalty term to the spline interpolation.

[0055] For thin-plate spline interpolation, terrain gradient constraints are incorporated into the classic thin-plate spline to improve the fitting accuracy for terrain undulation regions. The smoothing coefficient of the spline can also be adjusted in real time based on the terrain undulation provided by the digital elevation model: the smoothing coefficient is reduced for complex terrain to capture details of concentration changes, while it is appropriately increased for flat terrain to avoid overfitting.

[0056] In some embodiments, step S104 involves optimizing the UAV sampling path using a multi-objective optimization algorithm based on three-dimensional wind field scene data, digital elevation model data, and sampling point layout strategy to obtain a target three-dimensional flight path. This includes the step of optimizing the UAV sampling path using a non-dominated sorting genetic algorithm with spatial coverage, wind field characteristic error, and flight range energy consumption as optimization objectives to obtain a target three-dimensional flight path that satisfies the sampling point layout strategy.

[0057] Among them, a weighted graph structure can be constructed. ,in Indicates the sampling point. These are feasible path edges connecting the sampling points. Each edge weight The weighting comprehensively considers wind field compliance weight, elevation fluctuation penalty, and interpolation sensitivity weight. Wind field compliance weight. If the path direction aligns with the prevailing wind direction, energy consumption is reduced, and a lower weight is applied; if the path is against the wind or crosswind, the penalty coefficient is increased; elevation fluctuation penalty. When a path traverses terrain with significant undulations, energy consumption increases, thus increasing the cost of that path; interpolation sensitivity weight. Based on the sampling density requirements of the interpolation model for a specific area (e.g., Kriging interpolation in the main air duct area is more sensitive to point density), the path priority in that area should be appropriately increased.

[0058] Based on the above principles, the comprehensive path cost is defined as follows:

[0059] The default weight coefficient can be... It can be adaptively adjusted according to the scenario.

[0060] Next, based on this weighted graph model, a non-dominated sorting genetic algorithm (NSGA-II) is used for multi-objective path planning. The selected optimization algorithm is NSGA-II, whose parameters such as population size and number of generations can be set according to task requirements (e.g., population size 100, evolutionary iterations 50 generations). The algorithm's fitness function simultaneously evaluates the degree of monitoring requirement satisfaction and flight cost to search for the Pareto optimal set of routes. The main objectives of the comprehensive optimization include: maximizing spatial coverage, minimizing wind field characteristic errors, and optimizing flight range energy consumption.

[0061] Maximize spatial coverage: Improve the coverage of the monitoring area by the sampling path to ensure that there are sufficient sampling points at key locations and different height levels.

[0062] Minimizing wind field characteristic errors: To reduce the wind field representation error in areas not covered by the path, wind field heterogeneity indicators such as wind speed spatial variability rate and wind shear intensity are introduced to assess the uncertainty of sampling blind spots. The error factors are weighted by combining the terrain undulation provided by the digital elevation model, so that the path prioritizes passing through areas with drastic wind speed changes and complex terrain, thereby reducing the interpolation errors that may be generated in highly variable areas.

[0063] Flight range and energy consumption optimization: Control the total flight range and energy consumption of the UAV, and minimize the flight distance and reduce power consumption while meeting the requirements of sampling coverage and accuracy.

[0064] During path planning, constraints such as flight safety and communication reliability (e.g., obstacle avoidance safety distance, communication signal strength threshold, and remaining battery power limit) can be comprehensively considered to ensure the feasibility of the planning results. The three-dimensional sampled route obtained through evolutionary optimization has the ability to adapt to real-time wind direction changes: the path direction can be fine-tuned according to the current prevailing wind direction, reducing efficiency loss caused by headwind flight.

[0065] The optimization process is mainly based on the analysis of the objective function and constraints. The objective function is expressed as follows: f1 = 1 - Sampling point spatial coverage (maximization) f2 = Wind field characterization error (minimize) f3 = Drone energy consumption index (minimize) The constraints are expressed as follows: g1: Obstacle avoidance safety distance ≥ 10m g2: Communication link strength ≥ -90dBm g3: Remaining battery level ≥ 20% Under the aforementioned multi-objective and constraint conditions, a series of alternative Pareto optimal flight path schemes can be obtained through iterative evolution. The path with the best comprehensive index is then selected as the execution route based on decision preferences. This optimal route not only passes through all important sampling points but also undergoes fine-tuning and optimization based on real-time wind conditions, achieving environmental adaptability. For example, the path planning module can dynamically adjust the vertical spacing between adjacent flight paths based on the wind shear intensity at different altitude levels: at altitudes with high wind shear, the flight path spacing is appropriately reduced (approximately 30% less than the baseline planning spacing) to increase sampling density and capture concentration gradients caused by rapid wind field changes; while at altitudes with relatively stable wind fields, the flight path spacing is increased (approximately 30% more than the baseline) to reduce unnecessary overlapping coverage and improve cruising efficiency. Furthermore, to adapt to changes in the prevailing wind direction, the flight path is kept as consistent as possible with the current prevailing wind direction. The UAV can cruise back and forth along the prevailing wind direction or alternately sweep in a direction perpendicular to the prevailing wind direction. This path fine-tuning strategy ensures that the UAV flies with the wind as much as possible, reducing time and energy losses caused by headwinds, and more comprehensively capturing pollutant diffusion characteristics in both the downwind and crosswind directions.

[0066] In application, during the path execution phase, if the mission is interrupted or the environment changes abruptly, the above-mentioned scheme for real-time route planning can quickly respond and adjust the flight path. For example, when the prevailing wind direction suddenly changes, the ground control platform will replan the route for subsequent unsampled areas based on the latest wind field data, causing the UAV to turn to the new downwind direction to continue monitoring. If unforeseen dangerous situations are encountered during execution (such as the sudden appearance of restricted flight areas, severe weather, etc.), the system triggers an emergency reconfiguration mechanism, instructing the UAV to immediately change its route to avoid dangerous areas, or, if necessary, suspend the current mission and return to base. In addition, the ground platform can continuously monitor the UAV's status. When it detects that the UAV's remaining battery power is below the safety threshold or other signs of malfunction, it can control the UAV to adjust its route in advance and return directly to base, and can also dispatch a backup UAV to supplement sampling in areas not yet covered. By combining the above path planning with real-time adjustment strategies, the final output three-dimensional route comprehensively covers high-priority areas while being efficient and safe, ensuring that the UAV can complete the monitoring task with high quality.

[0067] In some embodiments, the three-dimensional pollutant monitoring method combined with wind field sensing further includes: evaluating the accuracy of the interpolation results using K-fold cross-validation and calculating error indices within the monitoring area; generating an error spatial distribution heatmap based on the error indices within the monitoring area; determining the area to be optimized based on the error spatial distribution heatmap; and determining a target optimization strategy based on the error level of the area to be optimized. The error indices include at least one of root mean square error, mean absolute error, and spatial consistency index; the optimization strategies include at least one of parameter-level optimization, structure-level optimization, and sampling-level optimization. Parameter-level optimization involves adjusting the parameter configuration of the current interpolation model according to preset rules; structure-level optimization involves switching the interpolation model based on error distribution characteristics; and sampling-level optimization involves planning additional UAV supplementary sampling routes for high-error areas.

[0068] The parameter-level optimization involves maintaining the current interpolation model structure while fine-tuning model parameters for regions with significant errors. For Kriging models, this includes refitting or replacing the semi-variogram (e.g., using a Gaussian model instead of an exponential model) and adjusting parameters such as the nugget effect, sill value, and correlation distance. For kernel density or spline models, parameters such as kernel bandwidth and spline smoothing coefficient are adjusted. Parameter adjustments refine the model's fit to existing data, reducing prediction bias. It's important to note that parameter adjustments can be predefined using rules, such as pre-setting adjustment steps for various parameters.

[0069] The structural level optimization involves: if the error does not significantly improve after parameter optimization, considering replacing the interpolation model or reconstructing the spatial correlation structure. By changing the basic structure and assumptions of the model, it is made more adaptable to the actual distribution characteristics of the observed data, thereby reducing systematic errors. The sampling level optimization involves: if local high-error areas still exist after adjusting model parameters and structure, a sampling supplementation mechanism is activated. The collaborative scheduling module will plan supplementary sampling flight paths based on the error spatial distribution heatmap: for example, for areas with prominent errors, a new local flight path is planned, and UAVs are dispatched to perform secondary sampling to obtain more real-time monitoring data for that area. The new sampled data will be incorporated into the interpolation calculation in real time to correct the previous model's bias in that area.

[0070] When using K-fold cross-validation to evaluate the accuracy of interpolation results, the original sampling data of the monitoring area is first randomly divided into K mutually exclusive subsets (e.g., K=5 or 10). Each time, one subset is selected as the validation set, and the remaining K-1 subsets are used as the training set. After interpolation calculation based on the training set, the measured values ​​of the validation set are compared with the interpolation results. This process is repeated K times, and the average of the evaluation results is taken as the final accuracy index. This reduces the impact of randomness in data partitioning on the evaluation. When generating an error spatial distribution heatmap based on these error indices, the error value of each grid cell needs to be associated with its corresponding spatial coordinates. A color gradient (e.g., from blue to red indicating from low to high error) visually presents the spatial distribution characteristics of the error. For example, high error areas may be concentrated in steep terrain or areas of sudden wind changes, while low error areas are mostly distributed in areas with gentle terrain and stable wind fields. When determining the area to be optimized based on the heatmap, continuous areas with error values ​​exceeding a preset threshold are prioritized, and then a target optimization strategy is formulated based on the error level. The error level can be determined based on the error level range in which the error value falls.

[0071] For example, for the mild error region, the corresponding optimization strategy can be parameter-level optimization. For the moderate error region, a combination of parameter-level and structure-level optimization can be used. For the severe error region, a sampling-level optimization strategy can be employed. Once the interpolation model achieves the expected accuracy through closed-loop optimization, a fully corrected three-dimensional pollutant concentration distribution map can be output, along with a detailed error assessment report. Simultaneously, based on the adjustment records generated during the optimization process, the parameter configuration of the interpolation algorithm and the sampling path scheme for the next UAV flight can be automatically updated, providing a basis for continuous monitoring improvement.

[0072] Based on the same inventive concept, such as Figure 4As shown, this application also provides a three-dimensional pollutant monitoring system 40 incorporating wind field sensing, comprising: a drone 41 and a processing module 42. The processing module 42 is communicatively connected to the drone 41, used to control the flight trajectory of the drone 41, and used to execute the three-dimensional pollutant monitoring method incorporating wind field sensing as described above. The processing module 42 can be deployed within the drone 41.

[0073] It should be noted that the pollutant three-dimensional monitoring system 40 combined with wind field perception provided in this embodiment and the pollutant three-dimensional monitoring method combined with wind field perception provided in this application embodiment are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned pollutant three-dimensional monitoring method combined with wind field perception, and the repeated parts will not be described again.

[0074] In some embodiments, such as Figure 5 As shown, the processing module 42 includes: a meteorological sensing module 421, a scheduling module 422, a processing module 423, and a communication module 424.

[0075] The meteorological sensing module 421 is used to collect meteorological data. The processing module 423 is connected to the meteorological sensing module 421 and is used to execute any of the above schemes for three-dimensional pollutant monitoring combined with wind field sensing. The scheduling module 422 is used to receive the target three-dimensional flight path output by the processing module 423 and coordinate multiple UAVs 41 to work collaboratively based on the target three-dimensional flight path, ensuring that the flight trajectories of each UAV 41 cover the target three-dimensional flight path. The communication module 424 is used to enable communication between the scheduling module 422 and each UAV 41.

[0076] It is understandable that the meteorological sensing module 421 can collect meteorological data of the monitoring area in real time, including key wind field parameters such as wind speed, wind direction, and turbulence intensity at different altitudes. The collection frequency can be dynamically adjusted according to monitoring needs (such as intensive collection under complex meteorological conditions), providing original meteorological basis for subsequent data processing and flight route planning.

[0077] After receiving real-time data transmitted by the meteorological sensing module 421, the processing module 423 can preprocess and classify the meteorological data according to wind field characteristics to obtain three-dimensional wind field scene data. Then, based on the three-dimensional wind field scene data and digital elevation model data, the processing module 423 can execute the aforementioned three-dimensional pollutant monitoring method combined with wind field sensing to obtain a target three-dimensional flight path that meets the requirements. After receiving the target three-dimensional flight path output by the processing module 423, the scheduling module 422 coordinates the division of labor among multiple UAVs 41 based on the spatial distribution characteristics of the flight path (such as flight path segmentation and key sampling point distribution) and the performance parameters of the UAVs 41 through task allocation algorithms (such as cluster-based segment allocation and load balancing scheduling), clarifying the responsible flight segment, take-off order and coordination rules of each UAV 41 (such as avoiding flight path intersections and ensuring no duplicate / omitted sampling point coverage), while simultaneously tracking the position and status of each UAV 41 in real time. The communication module 424 acts as an information interaction bridge between the scheduling module 422 and the UAVs 41, using low-latency, high-reliability wireless communication technologies (such as 4G / 5G and dedicated frequency bands for UAVs 41) to achieve bidirectional data transmission. The communication module 424 sends the route instructions and task adjustment information from the scheduling module 422 to each UAV 41. On the other hand, it receives the real-time position, remaining power, and collected data status information from the UAV 41, ensuring that the scheduling module 422 can dynamically correct the coordination strategy and ultimately ensure that multiple UAVs 41 can efficiently and collaboratively complete the sampling operation of the target three-dimensional route.

[0078] In some embodiments, the scheduling module 422 is used to divide the monitoring area into multiple sub-sampling areas according to the altitude layer, generate sub-flight routes corresponding to each sub-sampling area, and control each UAV 41 to perform cruise sampling along the sub-flight routes of each sub-sampling area.

[0079] It is understandable that by setting reasonable height intervals (such as every 50 meters, or adjusting the interval according to terrain undulations), each sub-sampling area focuses on environmental features within a specific vertical range. Simultaneously, by combining digital elevation model data to ensure that the bottom of each sub-sampling area is above the corresponding terrain at a safe height, flight conflicts of UAV 41 can be avoided. When generating sub-flight routes for corresponding sub-sampling areas, a path planning algorithm (such as the RRT algorithm) can be used to design continuous flight routes, combining the sampling point layout strategy of that layer, ensuring that the sub-flight routes cover all necessary sampling points in that layer, while optimizing the energy consumption and flight time of UAV 41. The scheduling module 422 coordinates the takeoff time, flight speed, and airspace avoidance rules of each UAV 41 through real-time communication, ultimately enabling multiple UAVs 41 to efficiently complete cruise sampling in their respective sub-sampling areas. This ensures that the three-dimensional environmental data of the entire monitoring area from low to high altitudes is collected completely and accurately, providing layered and refined basic data for subsequent data analysis.

[0080] In one example, the three-dimensional pollutant monitoring system 40, which incorporates wind field sensing, can achieve full automation from data acquisition to model generation through the collaborative work of its various modules, exhibiting good environmental adaptability and system scalability. The following describes the system's operation in conjunction with the task flow: 1. Task Issuance and Takeoff: The ground dispatch platform sets flight parameters (monitoring area range, target accuracy requirements, single-flight range limits, etc.) according to the monitoring task requirements and issues task instructions. The dispatch module 422 then calls the target 3D flight path generated in step S104 and assigns it to each UAV 41. If it is a single UAV 41, it executes the full-area flight path; if there are multiple UAVs 41, they execute their respective flight paths according to their assigned areas. Subsequently, the UAV 41 swarm takes off autonomously in sequence, each flying to its assigned sampling area and beginning cruise sampling along a predetermined optimized path.

[0081] 2. Data Acquisition and Transmission: During the flight of the UAV 41, onboard sensors collect meteorological data (from the meteorological sensing module 421, including wind speed, wind direction, turbulence intensity, etc.) and pollutant concentration data in real time, and continuously transmit the data back to the processing module 423 (e.g., a ground control platform) via a wireless communication network (such as a high-speed 5G link). The processing module 423 performs fusion processing on the real-time data, continuously updates the regional wind field model and pollutant distribution model, and incorporates the UAV 41's position, altitude, speed, battery level, and other status information into its monitoring to ensure flight safety.

[0082] 3. Dynamic Adjustment and Coordinated Response: If environmental conditions or flight conditions change during mission execution, the system will adjust its monitoring strategy in real time. Control commands will be sent through the scheduling module 422 to dynamically correct the flight paths and missions of the relevant UAVs 41, thereby maintaining monitoring effectiveness. For example, when a significant shift in the prevailing wind direction is detected, the processing module 423 will combine the latest wind field and current interpolation model prediction results to recalculate the cruise route for areas not yet monitored. It will then coordinate with the scheduling module 422 to direct the UAVs 41 to a new downwind area to continue tracking pollutant diffusion. In the event of sudden severe weather (such as heavy rain, strong winds, etc.) or the emergence of new obstacles / risk areas (such as temporary no-fly zones, toxic leak sources, etc.), the processing module 423 will coordinate with the scheduling module 422 to immediately adjust the flight altitude and path of the relevant UAVs 41, causing them to avoid dangerous areas. If necessary, the system will instruct the UAVs 41 to hover temporarily or return to standby until the environment improves before resuming the mission. When the scheduling module 422 detects that the communication signal of a certain UAV 41 is weakening or the positioning accuracy is decreasing, the scheduling module 422 will instruct the UAV 41 to adjust its flight attitude (such as increasing altitude) or change its flight route to improve communication quality, and notify other UAVs 41 to coordinate avoidance to reduce interference. If a UAV 41 needs to end its mission and return early due to insufficient power or malfunction, the scheduling module 422 will automatically replan the sampling task for the remaining area: other idle or group UAVs 41 can take over the patrol to supplement the sampling of the area not completed by the first UAV 41, or the uncompleted part can be included in the next round of mission plan. Through the classification response strategy for different environmental change factors, the system can ensure that even if the mission is interrupted or changes, it can quickly adjust and continue to complete the monitoring without missing any key data.

[0083] 4. Mission Completion and Data Output: When all planned sampling points have been covered or the UAV 41's battery level is close to the safety threshold and needs to return, the scheduling module 422 will terminate the current monitoring mission. Each UAV 41 will return sequentially according to predetermined instructions and automatically land at the takeoff base (UAV 41 nest) for charging or maintenance. After returning, the aircraft will upload complete flight logs, flight track records, and sampling data to the ground platform for further analysis and archiving. The processing module 423 executes step S105 based on the received data, ultimately reconstructing the three-dimensional pollutant concentration distribution model of the monitoring area, and can output a high-resolution three-dimensional pollutant concentration distribution map of the monitoring area, various statistical indicators, and detailed data reports. If the results report shows that the pollution status of certain areas or the regional model error reaches the threshold of concern (e.g., the regional pollutant concentration is greater than the set threshold or the model error is greater than the preset threshold), the processing module 423 will adjust the next monitoring strategy, and the scheduling module 422 will prioritize the flight of the corresponding area in subsequent missions to perform targeted supplementary monitoring of the area, entering a new closed-loop cycle. Through the aforementioned closed-loop operation mechanism, the entire system can self-update and improve after each round of monitoring, continuously enhancing the real-time performance, accuracy, and adaptability of UAV monitoring in complex environments, and providing an efficient and reliable technical means for three-dimensional monitoring of air pollutants in urban blocks, industrial parks, and other scenarios.

[0084] In some embodiments, an electronic device provided in this application includes a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the above-described three-dimensional pollutant monitoring method incorporating wind field sensing.

[0085] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.

[0086] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0087] This application also provides a computer-readable medium storing a computer program that, when executed by a processor, implements the aforementioned method for three-dimensional monitoring of pollutants incorporating wind field sensing. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into that device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.

[0088] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.

[0089] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A three-dimensional monitoring method for pollutants combining wind field sensing, characterized in that, include: Acquire three-dimensional wind field scene data and digital elevation model data of the monitoring area, and align the three-dimensional wind field scene data to the same spatial resolution as the digital elevation model data to obtain fused grid data of the monitoring area. The three-dimensional wind field scene data includes wind field structure sub-region information divided from the monitoring area. The sampling point layout strategy is determined based on the fused grid data, the preset wind field weight coefficient, and the preset terrain weight coefficient. The three-dimensional wind field scene data and the acquired sampling data are used as inputs to a pre-built matching model to obtain the interpolation strategy for the monitoring area. Based on 3D wind field scene data, digital elevation model data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the sampling path of the UAV, and obtain the target 3D flight path. The interpolation strategy is used to perform spatial interpolation on the target pollutant concentration data to reconstruct a three-dimensional pollutant concentration distribution model of the monitoring area. The target pollutant concentration data is the sampling data obtained by the UAV flying along the three-dimensional flight path of the target.

2. The three-dimensional pollutant monitoring method combining wind field sensing as described in claim 1, characterized in that, The acquisition of three-dimensional wind field scene data of the monitoring area includes: Acquire the collected multidimensional meteorological data; The multidimensional meteorological data is preprocessed to obtain wind field optimization data, wherein the preprocessing includes smoothing filtering and outlier removal; The optimized wind field data is input into a pre-constructed wind field structure recognition model to classify regional wind field features, thereby obtaining three-dimensional wind field scene data of the monitored area.

3. The three-dimensional pollutant monitoring method combining wind field sensing as described in claim 1, characterized in that, The step of determining the sampling point layout strategy based on the fused grid data, preset wind field weighting coefficient, and preset terrain weighting coefficient includes: Based on the fused grid data, preset wind field weighting coefficient, and preset terrain weighting coefficient, the sampling priority value of each grid point in the monitoring area is calculated to obtain a sampling priority matrix. The sampling point distribution set of each wind field structure sub-region is determined based on the fused grid data, the sampling priority matrix, and the sampling optimization strategy corresponding to each wind field structure label. The sampling point layout strategy for the monitoring area is determined by combining the sampling point distribution set of each wind field structure sub-region.

4. The three-dimensional pollutant monitoring method combining wind field sensing as described in claim 1, characterized in that, The wind field structure label corresponding to the wind field structure sub-region includes one of the following: downwind diffusion zone, shear variation zone, and static stagnation zone. The interpolation algorithm corresponding to the downwind diffusion zone is wind vector weighted kriging interpolation, wherein the wind vector weighted kriging interpolation algorithm is a kriging interpolation algorithm that introduces wind speed and wind direction as covariates into a semi-variogram model. The interpolation algorithm for the shear variation zone is an adaptive kernel density estimation algorithm, and the kernel function bandwidth of the adaptive kernel density estimation algorithm is inversely correlated with the degree of regional concentration change. The interpolation algorithm for the static and stagnant region is an improved thin-plate spline interpolation algorithm, which is a thin-plate spline interpolation algorithm that adds a terrain gradient penalty term to the spline interpolation.

5. The three-dimensional pollutant monitoring method combining wind field sensing as described in claim 1, characterized in that, Based on 3D wind field scene data, digital elevation model data, and sampling point layout strategy, a multi-objective optimization algorithm is used to optimize and plan the UAV sampling path, resulting in the target 3D flight path, including: With spatial coverage, wind field characteristic error, and range energy consumption as optimization objectives, a non-dominated sorting genetic algorithm is used to optimize and plan the sampling path of the UAV, thereby obtaining a target three-dimensional route that satisfies the sampling point layout strategy.

6. The three-dimensional pollutant monitoring method combining wind field sensing as described in claim 1, characterized in that, The method further includes: K-fold cross-validation is used to evaluate the accuracy of the interpolation results and to calculate the error index within the monitoring area. The error index includes at least one of the root mean square error, mean absolute error, and spatial consistency index. A spatial distribution heatmap of errors is generated based on the error indicators within the monitoring area; The optimization area is determined based on the error spatial distribution heatmap, and the target optimization strategy is determined based on the error level of the optimization area. The optimization strategy includes at least one of parameter-level optimization, structure-level optimization, and sampling-level optimization. The parameter-level optimization is to adjust the parameter configuration of the current interpolation model according to preset rules. The structure-level optimization is to switch the interpolation model according to the error distribution characteristics. The sampling-level optimization is to plan additional UAV supplementary sampling routes for high error areas.

7. The three-dimensional pollutant monitoring method combining wind field sensing as described in any one of claims 1 to 6, characterized in that, The wind field structure labels include the downwind diffusion zone, the shear variation zone, and the static stagnation zone; The sampling optimization strategy corresponding to each wind field structure label includes: increasing the density of sampling points along the main wind direction axis in the downwind diffusion zone by a first density increment value; In the critical region of wind speed gradient in the shear variation zone, the sampling coverage is increased with a second density increase value; in the static stagnant zone, the preset sampling density is maintained; in areas with severe topographic undulation, sampling points are set up in layers along the vertical direction to form a three-dimensional sampling network with multiple height layers.

8. A three-dimensional pollutant monitoring system incorporating wind field sensing, characterized in that, include: Drones and processing modules; The processing module is communicatively connected to the UAV and is used to control the flight trajectory of the UAV and to execute the three-dimensional pollutant monitoring method combined with wind field perception as described in any one of claims 1 to 7.

9. A three-dimensional pollutant monitoring system combining wind field sensing as described in claim 8, characterized in that, The processing module includes: a meteorological sensing module, a scheduling module, a processing module, and a communication module; The meteorological sensing module is used to collect meteorological data; The processing module is connected to the meteorological sensing module, and the processing module is used to execute the three-dimensional pollutant monitoring method combined with wind field sensing as described in any one of claims 1 to 7. The scheduling module is used to receive the target three-dimensional flight path output by the processing module, and coordinate multiple UAVs to work together based on the target three-dimensional flight path, so that the flight path of each UAV covers the target three-dimensional flight path. The communication module is used to enable communication between the scheduling module and each UAV.

10. A three-dimensional pollutant monitoring system combining wind field sensing as described in claim 9, characterized in that, The scheduling module is used to divide the monitoring area into multiple sub-sampling areas according to altitude, generate sub-flight routes corresponding to each sub-sampling area, and control each UAV to cruise and sample along the sub-flight routes of each sub-sampling area.

Images