Unmanned aerial vehicle monitoring network layout method based on mangrove forest ecological monitoring

Abstract

The application discloses a method for unmanned aerial vehicle monitoring network layout based on mangrove ecological monitoring, and relates to the technical field of unmanned aerial vehicle monitoring.The method acquires multi-source historical monitoring data to construct a spatiotemporal distribution atlas of mangrove ecological sensitivity, identifies the gradient change boundary of ecological risk grades and determines an initial layout density; collects meteorological and tidal prediction data to dynamically adjust the initial layout density in time sequence, and generates a dynamic layout density matrix; takes the matrix as a constraint, constructs a layout optimization function with the dual objectives of maximum monitoring coverage integrity and minimum monitoring energy consumption, and solves the function to obtain an optimal spatial coordinate set of unmanned aerial vehicle monitoring nodes; and generates a task flight path and a hovering point sequence of each node.The method realizes differential configuration of monitoring resources in the spatial dimension and dynamic adaptive adjustment in the time dimension, and takes into account the wide-area coverage integrity and low-energy-consumption demand of task execution.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) monitoring technology, specifically to a method for deploying a UAV monitoring network based on mangrove ecological monitoring. Background Technology

[0002] Mangrove wetland ecosystems possess extremely high ecosystem service value, and long-term, refined dynamic monitoring is fundamental for assessing ecological health and developing conservation strategies. Existing drone monitoring network deployment methods largely rely on uniform gridding or regular deployment based on constant density within the monitoring area. These methods ignore the spatial heterogeneity of ecological sensitivity within mangrove ecosystems. For example, vegetation cover, benthic habitat distribution, and tidal channel evolution trajectories vary significantly across different regions. Using a uniform monitoring density cannot allocate sufficient monitoring resources to ecologically fragile or highly sensitive areas. The main shortcomings of existing technologies are: firstly, the spatial configuration of drone monitoring nodes lacks a mechanism for linking it to the spatiotemporal dynamic characteristics of ecological risk, resulting in insufficient monitoring coverage in high-risk areas and redundant monitoring resources in low-risk areas, leading to a low overall cost-effectiveness ratio. Secondly, existing methods fail to incorporate dynamic environmental factors such as meteorology and tides into the network deployment decision-making process. Mangrove habitats are significantly affected by tidal fluctuations and wind field changes. Fixed monitoring node locations and task strategies can lead to coverage blind spots or a sharp increase in energy consumption under severe weather or special tidal conditions, failing to achieve adaptive matching between the monitoring scheme and the dynamic environment.

[0003] How to make the node layout of the monitoring network spatially adaptable to the gradient changes in ecological sensitivity, while responding to the dynamic disturbances of meteorological and tidal environments in time, is a problem that needs to be solved in the application of UAV ecological monitoring. In addition, how to coordinate the maximization of monitoring coverage and the minimization of the overall energy consumption of flight missions within a unified optimization framework is another problem that needs to be solved in building a UAV ecological monitoring network that can operate autonomously for a long time. Summary of the Invention

[0004] The purpose of this application is to provide a method for the layout of a drone monitoring network based on mangrove ecological monitoring, so as to introduce the spatial differences in mangrove ecological sensitivity and the temporal fluctuations of environmental conditions into the layout decision of the monitoring network, realize the on-demand differentiated configuration of monitoring nodes in the spatial dimension and the dynamic adaptive adjustment in the temporal dimension, and coordinate the completeness of monitoring coverage and the energy economy of task execution within a collaborative optimization model.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a method for deploying a drone monitoring network based on mangrove ecological monitoring, the method comprising:

[0006] Multi-source historical monitoring data of mangrove wetland ecosystems are acquired, and a spatiotemporal distribution map of mangrove ecological sensitivity is constructed based on this data. Preferably, satellite remote sensing historical image data, ground sensor historical monitoring data, and manual patrol historical data are sequentially processed by timestamp alignment and spatial coordinate normalization to obtain spatiotemporally aligned multi-source historical monitoring data. From this data, historical changes in mangrove vegetation cover, historical distribution ranges of benthic habitats, and historical evolution trajectories of tidal channels are extracted and input into a pre-trained ecological sensitivity assessment model to obtain an ecological sensitivity score for each spatial grid. This ecological sensitivity assessment model is preferably a spatiotemporal fusion model based on a long short-term memory network and a convolutional neural network. The ecological sensitivity scores of all spatial grids are then concatenated according to spatial coordinates and averaged along the time axis using a sliding window to generate a spatiotemporal distribution map of mangrove ecological sensitivity, accurately depicting the temporal and spatial variation patterns of mangrove ecosystem sensitivity.

[0007] Based on the spatiotemporal distribution map, the gradient change boundary of ecological risk level in the mangrove ecosystem is identified, and the initial deployment density of the UAV monitoring network in different ecological risk level areas is determined based on the gradient change boundary. As a technical solution of this invention, the ecological sensitivity score values ​​in the spatiotemporal distribution map are classified at equal intervals, and each spatial grid is assigned to its corresponding ecological risk level to obtain an ecological risk level distribution map. The differences in ecological risk levels between adjacent spatial grids are traversed, and when levels differ, their common boundary is marked as a gradient change boundary. The area corresponding to each ecological risk level is calculated, and the area is multiplied by a preset benchmark monitoring density coefficient to obtain the initial deployment density for each ecological risk level area. The benchmark monitoring density coefficient increases with the increase of ecological risk level, thereby deploying denser monitoring resources in high-risk ecological areas.

[0008] Meteorological and tidal forecast data of the mangrove monitoring area are collected, and the initial deployment density is dynamically adjusted according to the meteorological and tidal forecast data to generate a dynamic deployment density matrix for each monitoring period. In the preferred scheme, hourly wind speed forecast data, hourly rainfall probability forecast data, hourly high tide time data, and hourly tidal range data for the future monitoring cycle are collected from the meteorological forecast interface and the tidal forecast interface, respectively, and fused to obtain the comprehensive environmental disturbance vector for each monitoring period. The initial deployment density of each ecological risk level area and the comprehensive environmental disturbance vector of the corresponding period are input into a preset environmental attenuation function. The environmental attenuation function is preferably a multivariate nonlinear regression function with wind speed, rainfall probability, high tide time offset, and tidal range as input parameters. The dynamically adjusted density of each ecological risk level area in each monitoring period is calculated, and the dynamically adjusted densities of all areas are arranged according to spatial coordinates to generate a dynamic deployment density matrix, so that the monitoring density can be adaptively adjusted with changes in meteorological and tidal conditions.

[0009] Using a dynamic deployment density matrix as a constraint, a UAV monitoring network layout optimization function is constructed with the dual objectives of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. Solving this function yields the optimal set of spatial coordinates for UAV monitoring nodes. Specifically, the mangrove monitoring area can be discretized into a grid of candidate monitoring points. The mandatory selection probability weight of each candidate point is determined based on the dynamic deployment density matrix. The sub-objective function for maximizing monitoring coverage integrity is defined as the ratio of the union of the coverage areas of all selected UAV monitoring nodes to the total area of ​​the region, where the coverage area is determined by both the sensor detection radius and flight altitude. The sub-objective function for minimizing monitoring energy consumption is defined as the sum of the round-trip flight energy consumption and hovering energy consumption of all selected nodes. The mandatory selection probability weight is embedded as a constraint into the dual objective function, and a multi-objective particle swarm optimization algorithm is used for parallel iterative solution, outputting the optimal set of spatial coordinates corresponding to the optimal solution on the Pareto front. Preferably, the multi-objective particle swarm optimization algorithm introduces an adaptive inertia weight factor and a mutation operator, which can escape local optima in the later stages of iteration, balancing global search and local convergence accuracy. This optimization process effectively reduces the overall energy consumption of multi-drone collaborative monitoring while ensuring high coverage of key areas.

[0010] The task flight path and hovering point sequence for each UAV monitoring node are generated based on the optimal spatial coordinate set. As a further improvement of this invention, the three-dimensional coordinates of the hovering points of each UAV monitoring node are extracted from the optimal spatial coordinate set. Using the take-off and landing platform coordinates as the start and end points of the flight path, and all hovering points of that node as intermediate waypoints, an adaptive ant colony algorithm is used to plan the shortest closed flight path passing through all hovering points, outputting the task flight path for each node. The three-dimensional coordinates of the hovering points of each node are arranged according to the visiting order on the flight path to generate a hovering point sequence. Preferably, the adaptive ant colony algorithm uses a dynamically updated pheromone evaporation coefficient, which adaptively decreases as the remaining battery power of the UAV decreases, enabling path planning to dynamically respond to changes in endurance and reducing the risk of running out of power mid-flight.

[0011] After obtaining the optimal spatial coordinate set, the spatial location coordinates and monitoring capability parameters of existing fixed monitoring stations within the mangrove monitoring area can be acquired, and the existing monitoring coverage of the area by the existing fixed monitoring stations can be calculated. The spatial coordinates of each UAV monitoring node in the optimal spatial coordinate set are spatially overlaid with the existing monitoring coverage, and supplementary monitoring nodes located outside the existing coverage are marked. After these nodes are removed, the optimization function is re-solved to obtain the corrected optimal spatial coordinate set, thereby eliminating redundant coverage between the original fixed stations and saving UAV monitoring resources.

[0012] After generating the dynamic deployment density matrix, the phenological calendar of the breeding season of mangrove species in the mangrove monitoring area can be obtained to determine the ecologically sensitive time window. Within this time window, the dynamic deployment density matrix of the corresponding time period is read, and all density values ​​are amplified according to the preset sensitive time period multiplication factor to obtain the enhanced deployment density matrix. This matrix is ​​then used to replace the original time period matrix as the input constraint of the optimization function, thereby significantly enhancing the monitoring intensity during the ecologically sensitive period and capturing key phenological information.

[0013] After generating the mission flight path and hovering point sequence, the system can also collect real-time battery power data transmitted by each UAV monitoring node during mission execution. Combined with the distance between the current node and its take-off and landing platform, the maximum remaining cruise time can be calculated. When the maximum remaining cruise time of any node is lower than the preset safe return threshold, the system will assign the unvisited hovering point in the current hovering point sequence of that node to the nearest neighboring UAV monitoring node in terms of spatial distance, and update the mission flight path and hovering point sequence of that neighboring node. This will enable dynamic mission takeover, ensuring the safe return of the UAV and the continuous execution of the monitoring mission.

[0014] After obtaining the optimal spatial coordinate set, spatial distribution data and activity intensity timetables of human activity disturbance sources in the mangrove monitoring area can be acquired to calculate the human disturbance risk score for each spatial grid. This human disturbance risk score is then weighted and fused with the ecological sensitivity score in the spatiotemporal distribution map to generate a fused risk score distribution map. Based on the fused risk score distribution map, the drone monitoring nodes in the optimal spatial coordinate set are spatially clustered to identify risk hotspots. Additional drone monitoring node coordinates are added at their geometric centers to obtain an augmented optimal spatial coordinate set. This allows for targeted enhanced monitoring of areas with frequent human activities and high ecological sensitivity, improving the ability to perceive and respond to ecological disturbance events.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0016] Based on the spatiotemporal distribution map of mangrove ecological sensitivity, the gradient change boundary of ecological risk level is identified. This gradient change boundary serves as the basis for separating regions with different risk levels. Differentiated benchmark monitoring density coefficients are applied to regions with different ecological risk levels; the benchmark monitoring density coefficient increases with the risk level. This establishes a topological correlation between the initial deployment density and the spatial pattern of mangrove ecological sensitivity. This process results in a non-uniform spatial distribution of monitoring resources that matches the ecological risk gradient. High-risk areas receive denser monitoring node configurations, while redundant deployments in low-risk areas are suppressed. The spatial accuracy of monitoring resource deployment and its alignment with ecological protection needs are simultaneously improved. Meteorological and tidal forecast data for the monitoring area are collected and integrated into a comprehensive environmental disturbance vector for each monitoring period. Using an environmental attenuation function with wind speed, rainfall probability, high tide offset, and tidal range as input parameters, the initial deployment density for each ecological risk level region is dynamically adjusted time-by-time, generating a dynamic deployment density matrix. This matrix serves as a constraint for subsequent optimization, allowing the monitoring network layout to no longer rely on a fixed set of density parameters, but rather form a flexible deployment strategy that adjusts with time-varying meteorological and tidal environmental changes. When environmental conditions are harsh, the monitoring density can be adaptively reduced to avoid flight risks and energy waste, while when the environmental window is favorable, the monitoring density can be restored or enhanced to increase data acquisition intensity, thereby improving the mission robustness and environmental adaptability of the monitoring network.

[0017] A dual-objective optimization function is constructed using a dynamically deployed density matrix as a constraint. Maximizing monitoring coverage integrity and minimizing monitoring energy consumption are pursued as parallel optimization objectives, and a multi-objective particle swarm optimization algorithm is employed for iterative solution. This dual-objective architecture avoids the subjective weight bias introduced by a single weighted summation of coverage and energy consumption indicators, directly outputting the optimal set of node spatial coordinates on the Pareto front. This provides decision-makers with a network layout scheme that balances wide-area coverage and low-energy operation under a globally optimal trend, suppressing unnecessary energy consumption while pursuing full coverage. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of a method for deploying a drone monitoring network based on mangrove ecological monitoring;

[0020] Figure 2This is a flowchart illustrating the method for constructing a spatiotemporal distribution map of mangrove ecological sensitivity and determining the initial deployment density of a drone monitoring network.

[0021] Figure 3 It is a flowchart of the dynamic deployment density matrix generation based on meteorological and tidal forecasts;

[0022] Figure 4 This is a flowchart of the dual-objective optimization layout and flight path planning of UAV monitoring networks based on a dynamic deployment density matrix;

[0023] Figure 5 This is a flowchart of a method for optimizing the layout of a drone monitoring network that integrates existing fixed monitoring stations with ecologically sensitive windows.

[0024] Figure 6 It is a flowchart of the process of integrating risk score generation and monitoring node spatial clustering augmentation;

[0025] Figure 7 It is a time-series curve of the mangrove ecological sensitivity score;

[0026] Figure 8 These are curves showing the dynamic deployment density changes in areas with different ecological risk levels.

[0027] Figure 9 This is a schematic diagram of the Pareto front for multi-objective optimization of the UAV monitoring network layout;

[0028] Figure 10 This is a schematic diagram of the spatial layout of drone monitoring nodes and fixed monitoring stations in the mangrove monitoring area. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] See Figure 1This invention provides a method for deploying a drone monitoring network based on mangrove ecological monitoring. The method acquires multi-source historical monitoring data of the mangrove wetland ecosystem, constructs a spatiotemporal distribution map of mangrove ecological sensitivity, identifies the gradient change boundary of ecological risk levels based on the spatiotemporal distribution map, and determines the initial deployment density of the drone monitoring network in different ecological risk level areas. Then, it dynamically adjusts the initial deployment density in conjunction with meteorological and tidal forecast data to generate a dynamic deployment density matrix for each monitoring period. Next, using the dynamic deployment density matrix as a constraint, it constructs a drone monitoring network deployment optimization function with the dual objectives of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. Solving this optimization function yields the optimal spatial coordinate set of the drone monitoring nodes. Finally, it generates the mission flight path and hovering point sequence for each drone monitoring node based on the optimal spatial coordinate set.

[0031] Example 1:

[0032] In specific implementation, please refer to Figure 2 The steps for obtaining multi-source historical monitoring data of mangrove wetland ecosystems and constructing a spatiotemporal distribution map of mangrove ecological sensitivity based on multi-source historical monitoring data include data preparation, spatiotemporal alignment, feature extraction, model inference, and temporal smoothing.

[0033] Historical satellite remote sensing imagery, ground-based sensor monitoring data, and manual patrol records for the mangrove wetland ecosystem were utilized. The satellite remote sensing imagery data originated from long-term series images from multispectral satellites, with a spatial resolution better than 10 meters and a temporal resolution on a monthly scale. The ground-based sensor monitoring data came from soil salinity sensors, water level gauges, and water quality monitors deployed in the mangrove wetland ecosystem, with sampling frequency once per hour. The manual patrol records included species distribution quadrat survey records, disturbance event records, and visual assessment records of vegetation health.

[0034] The historical satellite remote sensing imagery, ground-based sensor monitoring data, and manual patrol records were processed with timestamp alignment and spatial coordinate normalization to obtain spatiotemporally aligned multi-source historical monitoring data. The timestamp alignment process involved aligning the acquisition time of the satellite remote sensing imagery to the most recent whole hour, converting the date in the manual patrol records to noon (12:00 PM) as the alignment timestamp, and directly resampling the ground-based sensor monitoring data to whole-hour times. The spatial coordinate normalization process involved converting the pixel coordinates of the satellite remote sensing imagery, the latitude and longitude coordinates of the ground-based sensor monitoring data, and the sample plot number coordinates of the manual patrol records to the UTM projection coordinate system, and then resampling them to a unified 5m × 5m spatial grid.

[0035] Historical changes in mangrove vegetation cover, historical distribution ranges of benthic habitats, and historical evolution trajectories of tidal canals were extracted from spatiotemporally aligned multi-source historical monitoring data. The extraction method for the historical changes in mangrove vegetation cover was as follows: vegetation cover was calculated for each uniform spatial raster using the Normalized Difference Vegetation Index (NDVI) thresholding method, resulting in vegetation cover values ​​for each uniform spatial raster at each aligned timestamp, thus forming the historical changes in mangrove vegetation cover. The extraction method for the historical distribution ranges of benthic habitats was as follows: spatial interpolation was used to spatially extend benthic habitat markers from ground-sensor historical monitoring data and manual patrol historical data, generating a benthic habitat distribution probability raster map at each aligned timestamp, thus forming the historical distribution ranges of benthic habitats. The extraction method for the historical evolution trajectories of tidal canals was as follows: water body indices from historical satellite remote sensing image data were thresholded and centerline extracted, resulting in tidal canal centerline vectors at each aligned timestamp, thus forming the historical evolution trajectories of tidal canals.

[0036] The historical change sequence of mangrove vegetation cover, the historical distribution range of benthic habitats, and the historical evolution trajectory of tidal channel systems are input into a pre-trained ecological sensitivity assessment model, which outputs an ecological sensitivity score for each uniform spatial grid in the mangrove wetland ecosystem. The ecological sensitivity assessment model employs a spatiotemporal fusion model based on a long short-term memory network and a convolutional neural network.

[0037] The network structure of the ecological sensitivity assessment model includes a spatial feature extraction branch, a temporal feature extraction branch, and a fusion output branch. The spatial feature extraction branch is a three-layer convolutional neural network. The first convolutional layer has 32 3×3 convolutional kernels with a stride of 1 and a modified linear unit (MRU) activation function; the second convolutional layer has 64 3×3 convolutional kernels with a stride of 1 and a MRU activation function; and the third convolutional layer has 128 3×3 convolutional kernels with a stride of 1 and a MRU activation function. Each convolutional layer is followed by a batch normalization layer and a 2×2 max-pooling layer. The temporal feature extraction branch is a two-layer long short-term memory (LSTM) network. The first LSM layer has 128 hidden units, the second LSM layer has 64 hidden units, and the time step unfolding length is 36 aligned timestamps. The fusion output branch concatenates the 128-dimensional spatial feature vector output from the spatial feature extraction branch and the 64-dimensional temporal feature vector output from the temporal feature extraction branch, and then passes it through a fully connected layer to output the ecological sensitivity score. The activation function of the fully connected layer is the Sigmoid function.

[0038] The training process of the ecological sensitivity assessment model is as follows: a training set is constructed using raster samples with historically labeled ecological sensitivity levels. Mean squared error is used as the loss function, and an adaptive moment estimation optimizer is employed. The initial learning rate is set to 0.001, the batch size to 64, and the number of training epochs to 200. Training is terminated early when the validation set loss no longer decreases. The input data is organized as follows: for each uniform spatial raster, the mangrove vegetation cover value, benthic habitat distribution probability value, and the nearest distance from the tidal channel midline to the center of the uniform spatial raster are taken from 36 consecutive aligned timestamps, forming a 36×3 time series matrix. Simultaneously, the above three feature values ​​from the most recent aligned timestamp of the uniform spatial raster and its eight surrounding uniform spatial raster areas are used to form a 3-channel spatial feature map input spatial feature extraction branch.

[0039] When reasoning with the ecological sensitivity assessment model, for each uniform spatial grid, the input data is organized in the manner described above, and the ecological sensitivity score is calculated through forward propagation. The ecological sensitivity score ranges from 0 to 1 as a continuous value.

[0040] The ecological sensitivity scores of all uniform spatial rasters are stitched together according to spatial coordinates to form an ecological sensitivity score raster image corresponding to each aligned timestamp. The stitched ecological sensitivity scores are then averaged along the time axis using a sliding window. The window width is set to 5 aligned timestamps, and the sliding step size is set to 1 aligned timestamp. Within each window, the arithmetic mean of the ecological sensitivity scores for all aligned timestamps is calculated. This average is assigned to the raster corresponding to the center aligned timestamp of the window. A symmetrical extension method is used at the boundaries to generate a spatiotemporal distribution map of mangrove ecological sensitivity. This spatiotemporal distribution map is a three-dimensional array with dimensions of the number of spatial horizontal coordinate rasters × the number of spatial vertical coordinate rasters × the number of time steps.

[0041] The specific steps for identifying the gradient change boundary of ecological risk level in mangrove ecosystems based on spatiotemporal distribution maps, and determining the initial deployment density of UAV monitoring network in different ecological risk level areas based on the gradient change boundary are as follows:

[0042] The ecological sensitivity scores in the spatiotemporal distribution map of mangrove ecological sensitivity were graded at equal intervals according to their numerical values. The ecological sensitivity scores were set to range from 0 to 1, and were divided into five equal intervals: [0, 0.2), [0.2, 0.4), [0.4, 0.6), [0.6, 0.8), and [0.8, 1.0]. The corresponding ecological risk levels, from low to high, were Level 1, Level 2, Level 3, Level 4, and Level 5. Each uniform spatial grid was assigned to its corresponding ecological risk level based on its ecological sensitivity score, resulting in an ecological risk level distribution map of the mangrove wetland ecosystem.

[0043] The ecological risk level differences of all adjacent uniform spatial graticles are traversed on the ecological risk level distribution map. Adjacent uniform spatial graticles refer to four-neighbor graticles that share a single edge. When two adjacent uniform spatial graticles have different ecological risk levels, their common boundary is marked as a gradient change boundary. The set of all marked gradient change boundaries forms the dividing line between regions with different ecological risk levels.

[0044] Calculate the area corresponding to each ecological risk level. For ecological risk level k, the area is... The calculation method is as follows: The number of uniform spatial grids with ecological risk level k in the ecological risk level distribution map is counted, multiplied by the area of ​​a single uniform spatial grid, where the area of ​​a single uniform spatial grid is 25 square meters. Based on the product of the area and the preset baseline monitoring density coefficient, the initial deployment density of the UAV monitoring network in each ecological risk level area is calculated using the following formula:

[0045]

[0046] in, Indicates the first The initial deployment density for each ecological risk level, in units of; Indicates the first The area of ​​each ecological risk level is expressed in square meters. Indicates the first The baseline monitoring density coefficient corresponding to each ecological risk level is expressed in units per square meter. It increases with the level of ecological risk. The value is preset as follows: corresponding to Level 1 risk Taking 0.0001, the corresponding level 2 risk Taking 0.0003, the corresponding level 3 risk Taking 0.0006, the corresponding level 4 risk Taking 0.0010, the corresponding level 5 risk The value is set to 0.0015. The above benchmark monitoring density coefficient is based on the following: at least 1.5 monitoring nodes are deployed per 1000 square meters in the highest risk area, and 1 monitoring node is deployed per 10000 square meters in the lowest risk area. Intermediate levels are determined by linear interpolation. The resulting initial deployment density... That is, in the first The total number of drone monitoring nodes that should be deployed in each ecological risk level area, and then... The monitoring node density value is evenly distributed among the uniform spatial grids within the area to obtain the initial monitoring node density value for each uniform spatial grid.

[0047] See Figure 7 The figure shows the changing trends of ecological sensitivity scores over time in mangrove wetland ecosystems with different ecological risk levels. The horizontal axis represents the aligned timestamp in months, covering approximately 100 months, while the vertical axis represents the ecological sensitivity score, ranging from 0 to 1. The figure contains five curves, corresponding to ecological sensitivity scores for risk levels one through five, distinguished by different symbols: level one risk is represented by a solid circle, level two by a solid square, level three by a solid triangle, level four by a solid rhombus, and level five by a solid asterisk.

[0048] As can be seen from the trend of the curves, the ecological sensitivity scores of the Level 5 risk areas are generally at the highest level, fluctuating between approximately 0.85 and 0.95, indicating that the ecological sensitivity of these areas has remained at a high level for a long time, with some temporal fluctuations. The scores of the Level 4 risk areas are next, mainly fluctuating between 0.63 and 0.77, showing a periodic trend with alternating peaks and troughs. The ecological sensitivity scores of the Level 3 risk areas are approximately between 0.43 and 0.57, with significant fluctuations and relatively pronounced periodic changes. The scores of the Level 2 risk areas are between 0.25 and 0.35, with smaller fluctuations and relative stability. The scores of the Level 1 risk areas are the lowest, mainly distributed in the range of 0.08 to 0.15, and are generally stable.

[0049] Each risk level curve exhibits a certain periodic fluctuation characteristic, especially in the level 3 and 4 risk areas, with a cycle of approximately 20 months, reflecting the spatiotemporal dynamic changes in the ecological sensitivity of mangroves. The clear stratification of ecological sensitivity scores between levels is consistent with the ecological risk level classification standard set in Patent Example 1, that is, the five risk levels are divided by the numerical range of ecological sensitivity scores, with higher scores corresponding to higher risk levels.

[0050] Overall, the figure clearly shows the changes in the ecological sensitivity scores of different ecological risk levels of mangrove wetlands based on the spatiotemporal distribution map over a long-term monitoring time series, providing data basis for determining the initial deployment density and dynamic adjustment of the UAV monitoring network based on this score.

[0051] Example 2:

[0052] In specific implementation, please refer to Figure 3 The process of collecting meteorological and tidal forecast data of the mangrove monitoring area, dynamically adjusting the initial deployment density based on the meteorological and tidal forecast data, and generating a dynamic deployment density matrix for each monitoring period includes meteorological and tidal forecast data collection, fusion of comprehensive environmental disturbance vectors, construction and training of environmental attenuation functions, calculation of dynamically adjusted density, and generation of dynamic deployment density matrix.

[0053] Hourly wind speed forecast data, hourly precipitation probability forecast data, hourly high tide time data, and hourly tidal range data for the mangrove monitoring area within the future monitoring period are collected from the meteorological forecast interface and the tidal forecast interface, respectively. The future monitoring period is set to 120 hours from the current time. The meteorological forecast interface accesses the Global Forecast System (GFS) data provided by the Numerical Weather Prediction Center, with a spatial resolution of 0.25 degrees and a temporal resolution of 3 hours. The meteorological forecast data is downscaled to the center point of the mangrove monitoring area through bilinear interpolation, and then converted into an hourly sequence through temporal linear interpolation. The wind speed component at a height of 10 meters is extracted and synthesized into a scalar wind speed to obtain hourly wind speed forecast data in meters per second. At the same time, the total precipitation probability forecast field is extracted to obtain hourly precipitation probability forecast data, with values ​​ranging from 0 to 1 decimals. The tidal forecast interface accesses the harmonic constant forecast product output by the regional tidal model, reads the forecast results of the tide gauge stations in the mangrove monitoring area, obtains the tidal height value at each hour in the next 120 hours, obtains hourly high tide time data by identifying the extreme points of the tidal curve, and marks the hourly time with Coordinated Universal Time, as well as hourly tidal range data, in meters. Both the hourly high tide time data and the hourly tidal range data are arranged in the forecast time sequence.

[0054] Hourly wind speed forecast data, hourly rainfall probability forecast data, hourly high tide time data, and hourly tidal range data are merged using the same timestamp. The monitoring period is measured in whole hours. For each monitoring period, extract the corresponding hourly wind speed forecast data. Hourly rainfall probability forecast data Offset corresponding to hourly high tide time data and hourly tidal range data These form a four-dimensional vector, defined as the comprehensive environmental disturbance vector. , Hourly high tide time offset The calculation method is as follows: obtain the predicted highest tide time within the calendar day of the monitoring period. ,calculate The unit is hours; if there is no climax event during that calendar day, then Take a fixed value of 24.

[0055] The environmental degradation function employs a multivariate nonlinear regression model, specifically a four-layer fully connected neural network. The network input layer contains four neurons, receiving the combined environmental perturbation vector. The input layer has four components. The first hidden layer contains 16 neurons, fully connected to the input layer, and uses the ReLU activation function. The second hidden layer contains 8 neurons, fully connected to the first hidden layer, and uses the ReLU activation function. The output layer contains 1 neuron, fully connected to the second hidden layer, and uses the sigmoid function to output a scalar decay coefficient with a range of [value missing]. Environmental attenuation function The mathematical expression is:

[0056]

[0057] in, Indicates the first The comprehensive environmental disturbance vector for each monitoring period has the following dimensions: ; The weight matrix from the input layer to the first hidden layer has a dimension of . ; Let be the bias vector of the first hidden layer, with dimension . ; The weight matrix from the first hidden layer to the second hidden layer has a dimension of . ; Let be the bias vector of the second hidden layer, with dimension . ; This is the weight matrix from the second hidden layer to the output layer, with dimension 1. ; This is the bias scalar of the output layer; It is a linear rectifier function, defined as taking the larger value between the input value and 0 element by element; The Sigmoid function is defined as follows: .

[0058] The training process for the environmental attenuation function is as follows: Historical meteorological reanalysis data and historical tidal measurement data for the mangrove monitoring area over the past three years are collected hourly. A comprehensive environmental disturbance vector is constructed as the training input sample in the manner described above. Simultaneously, the ratio of the actual effective monitoring coverage area to the interference-free theoretical coverage area recorded in the corresponding hourly UAV test flights is collected as the attenuation coefficient label value. The sample set is randomly divided into a training set and a validation set in a 7:3 ratio, and the input features are normalized. Divide by the historical maximum observed wind speed of 30 m / s, Keep the original value. Divide by 24, Divide by the historical maximum tidal range of 10 meters. Use mean squared error as the loss function, and employ an adaptive moment estimator (AME) as the optimization algorithm. Set the learning rate to 0.0005, batch size to 32, and maximum training epochs to 500. Calculate the validation set loss after each training epoch. If the validation set loss does not decrease within 20 consecutive epochs, trigger early stopping and save the model parameters with the minimum validation set loss, fixing the weight matrix. , , and bias vector , and bias scalar The specific value.

[0059] For each area with an ecological risk level, the ecological risk level will be determined. Corresponding initial deployment density With the The comprehensive environmental disturbance vector for each monitoring period Input to the environment decay function after training is complete In the middle, calculate the density after dynamic adjustment. The calculation formula is:

[0060]

[0061] in, Indicates the first The ecological risk level area is in the first The density after dynamic adjustment for each monitoring period. The first step determined in the preceding steps The initial deployment density for each ecological risk level, This is the dimensionless attenuation coefficient output by the environmental attenuation function. A value closer to 1 indicates a less significant impact of environmental conditions on the attenuation of the monitored data. (The last sentence appears to be incomplete and possibly refers to a different context.) Calculated from all ecological risk level areas within the region The grid is filled according to the spatial coordinates of the corresponding area. Each unified spatial grid inherits the dynamically adjusted density value of its corresponding ecological risk level area, forming a two-dimensional matrix. The row index of the matrix corresponds to the spatial horizontal coordinate, the column index corresponds to the spatial vertical coordinate, and the element value is the dynamic monitoring density value of that unified spatial grid. This two-dimensional matrix is ​​the [missing information - likely a specific type of matrix]. The dynamic deployment density matrix for each monitoring period is generated. The above operation is performed for each future monitoring period to generate a sequence of dynamic deployment density matrices for all monitoring periods.

[0062] See Figure 8 The horizontal axis of the graph represents the monitoring period in hours, ranging from 0 to 120 hours, while the vertical axis represents the dynamically adjusted monitoring density in units. The graph shows the dynamic adjustment monitoring density changes for five ecological risk level areas (Level 1, Level 2, Level 3, Level 4, and Level 5) during different monitoring periods. Each curve is distinguished by a different line type and marker: Level 1 risk is represented by a solid dot, Level 2 risk by a dashed square, Level 3 risk by a dotted triangle, Level 4 risk by a dashed diamond, and Level 5 risk by a star-shaped dashed line.

[0063] The monitoring density after dynamic adjustment is highest in Level 5 risk areas, fluctuating between 7 and 12 units, indicating significant volatility and reflecting the strong impact of environmental disturbances on the monitoring needs of areas with the highest ecological risk level. Level 4 risk areas have the second highest monitoring density, with fluctuations mainly between 4 and 8 units. Level 3 risk areas show relatively gentle fluctuations, with density values ​​between 2.5 and 4.5 units. The monitoring densities after dynamic adjustment are lower and fluctuate less, remaining between approximately 1 to 2.5 units and 0.3 to 1 unit, respectively.

[0064] The overall curves exhibit distinct periodic fluctuations, indicating that environmental disturbances reflected in meteorological and tidal forecast data have a temporal dynamic adjustment effect on monitoring density in areas with different ecological risk levels. Peak monitoring density occurred in multiple monitoring periods, particularly around the 5th, 55th, 75th, and 100th hours. The peak monitoring density in level 5 and level 4 risk areas was especially significant, suggesting that environmental conditions had a substantial impact on the mangrove ecosystem during these periods, necessitating increased deployment density of the UAV monitoring network to address potential ecological risks.

[0065] As can be seen from the figure, the technical solution in Example 2, which uses meteorological and tidal forecast data to dynamically adjust the initial deployment density through an environmental attenuation function to generate a dynamic deployment density matrix, effectively achieves dynamic adaptation of monitoring density in areas with different ecological risk levels, and ensures reasonable configuration and resource optimization of the UAV monitoring network as environmental conditions change.

[0066] Example 3:

[0067] In specific implementation, please refer to Figure 4 Using a dynamic deployment density matrix as a constraint, a UAV monitoring network layout optimization function is constructed with the dual objectives of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. Solving the UAV monitoring network layout optimization function yields the optimal set of spatial coordinates for UAV monitoring nodes.

[0068] The mangrove monitoring area was discretized into a grid of candidate monitoring points. The candidate monitoring point grid was generated as follows: using the coverage area of ​​the mangrove monitoring area as the boundary, the three-dimensional coordinates of the center point of a uniform spatial grid were used as candidate monitoring points. The elevation coordinates were extracted from the airspace location of the UAV flight altitude layer from the Digital Land Model (DSM), with the flight altitude layer set at 80 meters above sea level. The total number of candidate monitoring points is denoted as […]. Each candidate monitoring point has a unique number. , .

[0069] The required probability weight for each candidate monitoring point is determined based on the dynamic deployment density matrix. For each candidate monitoring point... Read candidate monitoring points The unified spatial grid is in the first The dynamic monitoring density value in the dynamic deployment density matrix for each monitoring period is denoted as: .Will Normalization process yields the required probability weights. The normalization method is , This is the sum of the dynamic monitoring density values ​​corresponding to all candidate monitoring points. (Required probability weight) The range of values ​​is Characterize candidate monitoring points In the The degree of preference for selecting a monitoring period as a drone monitoring node.

[0070] Define a sub-objective function that maximizes monitoring coverage integrity. This is the ratio of the union of the coverage areas of all selected drone monitoring nodes to the total area of ​​the mangrove monitoring area. Decision variable. For a length of binary vector, ,in, A value of 1 indicates a candidate monitoring point. Selected as a drone monitoring node A value of 0 indicates that the node is not selected. The coverage area of ​​each UAV monitoring node is determined by the detection radius of its configured onboard multispectral sensor. and flight altitude It was jointly decided that the ground coverage area would be an area centered on the vertical projection of the hovering point, with a radius of [missing information]. The circular area , The sensor's nominal height is 100 meters. The sub-objective function for maximizing monitoring coverage integrity is... ,in Indicates the selected node The circular area covered This represents the total area of ​​the mangrove monitoring area. This is the area calculation function.

[0071] Define a sub-objective function to minimize energy consumption monitoring. This is the sum of the flight energy consumption of all selected UAV monitoring nodes from the take-off and landing platform to their hovering point, and the hovering energy consumption of each node while performing monitoring tasks at the hovering point. It is also the one-way flight energy consumption of the UAV monitoring node from the take-off and landing platform to the hovering point. Flight distance A linear relationship exists. , The energy consumption coefficient per unit flight distance for the drone is set to 0.5 Wh / m. To determine the candidate monitoring points from the three-dimensional coordinates of the take-off and landing platform The Euclidean distance between the three-dimensional coordinates. Hovering energy consumption. , The hovering energy consumption coefficient for the drone is set to 200 watt-hours per unit time. The preset monitoring duration at a single hovering point is set to 0.25 hours. Therefore, the sub-objective function for minimizing monitoring energy consumption is... .

[0072] The required probability weights are embedded as constraints into the sub-objective functions of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. Specifically, in the multi-objective optimization process, only when the randomly generated candidate solutions satisfy the condition... The node's required probability weight Exceeding the preset probability threshold Only when this condition is met is the solution accepted into the non-dominated solution set; a preset probability threshold is used. The value is set to 0.01. Furthermore, the mandatory probability weights are integrated into the particle swarm initialization process, with the initial values ​​of each dimension of the position vector of each particle set as probabilities. Choose 1, with probability Take 0.

[0073] A multi-objective particle swarm optimization algorithm is used to maximize the sub-objective function of monitoring coverage integrity. and monitoring energy consumption minimization sub-objective function Parallel iterative solutions are performed. In the multi-objective particle swarm optimization algorithm, the particle swarm size is set to 100 particles, and the position vector of each particle represents a binary decision vector. The velocity vector has the same dimension as the position vector, and the velocity components take continuous values. Particle positions are initialized using the aforementioned mandatory probability weighting method. The algorithm iterates 200 times, calculating two objective function values ​​for each particle in each generation. and The individual optimal position and global optimal position set of each particle are updated according to the Pareto dominance relationship. Non-dominated solutions are stored in an external archive with a capacity of 50. When the archive is full, crowding distance sorting is used for pruning.

[0074] Introducing an adaptive inertia weight factor in multi-objective particle swarm optimization algorithms And mutation operator. Adaptive inertia weighting factor. It decreases linearly with the number of iterations. ,in, Take 0.9, The value is set to 0.4, where iter is the current iteration number and maxiter is the maximum iteration number of 200. Velocity updates use a binary particle swarm optimization formula, and a mutation operator is added during position updates, with mutation probability... Adaptive adjustment based on particle crowding distance in each iteration: for particles with crowding distance below the average, For particles with a crowding distance higher than average, The mutation operation randomly flips the binary value of one dimension of the position vector. Finally, a point at a distance from the ideal point is selected from the Pareto front of the external archive. The solution with the minimum Euclidean distance is considered the optimal compromise solution, and the binary vector corresponding to this solution is... This is the optimal set of spatial coordinates for the UAV monitoring nodes.

[0075] The process of generating the mission flight path and hovering point sequence for each UAV monitoring node based on the optimal set of spatial coordinates is as follows.

[0076] Extract the three-dimensional coordinates of the hovering point of each UAV monitoring node from the optimal set of spatial coordinates. The candidate monitoring point numbers with a median value of 1 are extracted, and the three-dimensional coordinates of the corresponding candidate monitoring points are read, which are the three-dimensional coordinates of the hovering points of the UAV monitoring nodes. At the same time, the coordinates of the take-off and landing platform of each UAV monitoring node are obtained as the starting coordinates and ending coordinates of the flight path. The take-off and landing platform is fixedly set in an open area of ​​land at the edge of the mangrove monitoring area, and the coordinates are pre-measured values.

[0077] For each UAV monitoring node, using the starting point coordinates as the starting point, all hovering points' 3D coordinates as intermediate points, and the ending point coordinates as the ending point, an adaptive ant colony algorithm is used to plan the shortest closed route passing through all hovering points' 3D coordinates, outputting the task route for each UAV monitoring node. In the adaptive ant colony algorithm, the number of ants is set to 30, and the number of iterations is set to 100 generations. Each ant starts from the take-off and landing platform, and at each step, uses a pseudo-random proportional rule to select the next hovering point to be visited. The selection probability is determined by the product of pheromone concentration and a heuristic factor. The heuristic factor... Take as node To the node The reciprocal of the Euclidean distance between them.

[0078] The adaptive ant colony algorithm uses a dynamically updated pheromone evaporation coefficient. pheromone volatility coefficient The remaining power of the monitoring node is monitored by the drone. The dynamic adjustment formula is: [The formula is missing from the original text, but it seems to be missing from the original text.]

[0079]

[0080] in, The pheromone volatility coefficient currently in use; This is the upper bound of the pheromone evaporation coefficient, with a value of 0.3. This is the lower bound of the pheromone evaporation coefficient, with a value of 0.05. The full charge value for the drone monitoring node is set to 500 watt-hours. The current remaining battery power is estimated linearly based on the length of the planned flight segment, with an initial value of [value missing]. After each segment of the route, from After deducting flight energy consumption, the calculation method for flight energy consumption is the same as described above. Consistent. The pheromone update rule is: after each iteration, a global update is performed on the pheromones along the path. upper pheromone concentration according to Update, in which, The pheromone increment of the ant along the path is inversely proportional to the total length of the path traveled by the ant.

[0081] Update after each iteration The corresponding simulated voltage value was used to recalculate the pheromone evaporation coefficient. This process slows down pheromone evaporation as remaining battery power decreases, thus enhancing the retention of existing, better paths. After the iterations are completed, the shortest closed-loop route in all iterations is selected as the mission route for the UAV monitoring node.

[0082] The three-dimensional coordinates of all hovering points of each UAV monitoring node are arranged according to the order of access on the mission route to generate a hovering point sequence for each UAV monitoring node. The hovering point sequence is directly extracted from the intermediate points of the planned mission route and output as a list of sequences other than the take-off and landing platform as the start and end points of the sequence.

[0083] Example 4:

[0084] In specific implementation, please refer to Figure 5 After solving the optimization function of the UAV monitoring network layout to obtain the optimal set of spatial coordinates of the UAV monitoring nodes, the spatial location coordinates and monitoring capability parameters of the existing fixed monitoring stations in the mangrove monitoring area are obtained. Based on the spatial location coordinates and monitoring capability parameters of the existing fixed monitoring stations, the existing monitoring coverage of the existing fixed monitoring stations in the mangrove monitoring area is calculated.

[0085] Existing fixed monitoring stations include ground-based meteorological stations, hydrological monitoring towers, and automatic water quality monitoring buoys permanently deployed within the mangrove wetland ecosystem. Spatial location coordinates are the three-dimensional coordinates of each fixed monitoring station in the UTM projection coordinate system. Monitoring capability parameters include the effective detection radius of the sensors at each fixed monitoring station. And the effective monitoring height range, including the sensor's effective detection radius. The detection range is determined based on the maximum detection distance provided in the technical specifications of the fixed monitoring station equipment, in meters. The existing monitoring coverage area of ​​each existing fixed monitoring station is defined as a circle centered on the spatial coordinates of the fixed monitoring station and extending to the effective detection radius of the sensor. Let be the projection of a circular region with radius on the horizontal plane. When the sensor mounted on the fixed monitoring station has height constraints, the existing monitoring coverage area is the effective detection radius of the sensor. The conical projection area is determined together with the sensor's pitch angle.

[0086] The spatial coordinates of each UAV monitoring node in the optimal spatial coordinate set are spatially overlaid with the existing monitoring coverage area to identify UAV monitoring nodes located outside the existing monitoring coverage area as supplementary monitoring nodes. The spatial overlay analysis process involves iterating through the 3D coordinates of the hovering point of each UAV monitoring node in the optimal spatial coordinate set, and then comparing the horizontal projection coordinates of each UAV monitoring node with the circular areas of the existing monitoring coverage area of ​​all fixed monitoring stations to determine point inclusion. If the horizontal projection coordinates of a UAV monitoring node fall within any circular area of ​​the existing monitoring coverage area, then the UAV monitoring node is covered by the existing monitoring coverage area; if the horizontal projection coordinates of a UAV monitoring node do not fall within any circular area of ​​the existing monitoring coverage area, then the UAV monitoring node is marked as a supplementary monitoring node.

[0087] After removing the spatial coordinates of the supplementary monitoring nodes from the optimal spatial coordinate set, the solution process for the UAV monitoring network layout optimization function is re-executed. The removal operation involves generating a new set of candidate monitoring points, deleting the candidate monitoring points occupied by the supplementary monitoring nodes from the original candidate monitoring point grid, and leaving the remaining candidate monitoring points unchanged. Using the removed candidate monitoring point set as input, the multi-objective particle swarm optimization algorithm with dynamic deployment density matrix as constraint and maximizing monitoring coverage integrity and minimizing monitoring energy consumption as dual objectives is re-executed to obtain the corrected optimal spatial coordinate set. The corrected optimal spatial coordinate set, together with the candidate monitoring points that were not removed and the existing fixed monitoring stations, constitutes the complete monitoring network node layout of the mangrove monitoring area.

[0088] In practice, after generating a dynamic deployment density matrix for each monitoring period, the phenological calendar of the breeding season of mangrove species in the mangrove monitoring area is obtained, and the ecologically sensitive time window is determined based on the phenological calendar of the breeding season.

[0089] The reproductive phenological calendar for mangrove species was obtained by querying plant phenological observation databases at the latitude of the mangrove wetland ecosystem. The calendar records the start and end dates of the flowering, embryonic development, and viviparous seedling shedding periods of dominant mangrove species. The ecologically highly sensitive time window is defined as the continuous period from 7 days before the start date of the flowering period for all species in the reproductive phenological calendar to 7 days after the end date of the viviparous seedling shedding period. If the reproductive periods of multiple species overlap, the overlapping periods are merged into a single continuous ecologically highly sensitive time window.

[0090] Within the ecologically sensitive time window, the dynamic deployment density matrix for the corresponding monitoring period is read, and all density values ​​in the dynamic deployment density matrix are amplified according to a preset sensitivity period multiplication factor to obtain the enhanced deployment density matrix within the ecologically sensitive time window. (Preset sensitivity period multiplication factor) The method for determining it is as follows: The value was set at 1.8. This value is based on the fact that during the mangrove species' breeding season, the marginal impact of human activity on reproductive success is approximately 1.8 times that of the non-breeding season. Therefore, the monitoring density was increased by 1.8 times to maintain the same level of monitoring coverage integrity. This applies to each monitoring period within the ecologically sensitive time window. Read the dynamic layout density matrix , Each element in the matrix represents the dynamic monitoring density value corresponding to a uniform spatial grid, and the enhanced deployment density matrix is ​​calculated. .

[0091] To enhance the deployment density matrix Replace the dynamic deployment density matrix of the original monitoring period As an input constraint to the UAV monitoring network layout optimization function, the mandatory probability weights of the multi-objective particle swarm optimization algorithm in subsequent steps will be based on... The values ​​of each element are recalculated, which allows the deployment of drone monitoring nodes to concentrate in highly sensitive areas during the ecologically sensitive time window.

[0092] See Figure 9 The figure illustrates the spatial distribution of the UAV monitoring network and the coverage area of ​​existing fixed monitoring stations within the mangrove monitoring area. The horizontal axis represents the UTM projection, ranging from approximately 400,000 to 402,000 meters, while the vertical axis represents the UTM projection, ranging from approximately 2,000,000 to 2,003,000 meters. The legend identifies three types of key locations: candidate monitoring points are represented by small gray dots, scattered throughout the monitoring area, numerous and evenly distributed; UAV monitoring nodes are marked with hollow circles, sparsely distributed and covering the candidate points; supplementary monitoring nodes are marked with diamonds, mainly concentrated at the edges of the existing fixed monitoring station coverage area and in some blank areas, serving to fill monitoring coverage gaps. Fixed monitoring stations are represented by triangles. The figure shows four fixed stations, each surrounded by a black dashed circle indicating its monitoring coverage area. The circle radius corresponds to the effective detection radius of the sensor, and the coverage area size is relatively fixed. The observation shows that the coverage areas of the fixed monitoring stations in the diagram partially overlap, and some UAV monitoring nodes are clearly located outside the dashed circles, indicating that these nodes supplement the blind spots of the fixed stations. The spatial distribution of the UAV monitoring nodes is relatively uniform, with deployments in both highly sensitive and peripheral areas, indicating that the optimization algorithm fully considers the dynamic deployment density matrix and the trade-off between coverage integrity and energy consumption. The layout of the supplementary monitoring nodes reflects the steps in Example 4 of optimizing the layout after removing redundant nodes within the existing monitoring coverage area through spatial overlay analysis, ensuring the rational use of monitoring resources and the integrity of the coverage network. The overall layout fully reflects the optimization design concept of the UAV monitoring network based on mangrove ecological monitoring.

[0093] Example 5:

[0094] In specific implementation, please refer to Figure 6After generating the mission flight path and hovering point sequence for each UAV monitoring node based on the optimal spatial coordinate set, the system collects real-time battery power data transmitted by each UAV monitoring node during the execution of its mission flight path. Each UAV monitoring node is equipped with a battery power monitoring module, which collects real-time battery voltage and current at a sampling frequency of 1 Hz. The battery power data is compressed and packaged into packets every 10 seconds via an onboard 4G communication module and transmitted back to the ground control station. Each data packet contains a unique identifier for the UAV monitoring node, a collection timestamp, and the current remaining battery watt-hours. After parsing the battery power data, the ground control station obtains the current remaining battery watt-hours for each UAV monitoring node, denoted as... The unit is watt-hour.

[0095] Based on battery level data and the distance between the current drone monitoring node and its takeoff and landing platform, calculate the maximum remaining cruise time for each drone monitoring node. (Distance between the drone monitoring node and its takeoff and landing platform) Defined as the Euclidean distance between the three-dimensional coordinates of the current hovering point of the UAV monitoring node and the three-dimensional coordinates of the take-off and landing platform, in meters. Maximum remaining cruise time. The calculation method is as follows: ,in, The unit time flight power of the UAV monitoring node is taken as 800 watt-hours. The hovering power per unit time for the drone monitoring node is set to 200 watt-hours. Represents the average power consumption rate of the UAV monitoring node during flight and monitoring missions, expressed in watt-hours per hour. Maximum remaining cruise time. The unit is hours, representing the total time that the drone monitoring node can continue to fly and hover under the current battery level.

[0096] When the maximum remaining cruise time of a certain drone monitoring node Below the preset safe return threshold At this time, the task reassignment process is triggered. Safe return threshold. The value is 0.5 hours, which is the safe return threshold. The 0.5-hour time limit is based on the fact that the maximum flight time required for a UAV monitoring node to return from the deepest part of the mangrove monitoring area to the take-off and landing platform is statistically no more than 0.3 hours, with a 0.2-hour safety margin reserved to cope with sudden airflow disturbances. All unvisited hovering points in the current hovering point sequence of the UAV monitoring node are marked as hovering points to be assigned. Each hovering point to be assigned is sequentially retrieved from the list of unvisited hovering points, and the Euclidean spatial distance between the hovering point to be assigned and the current location of all other UAV monitoring nodes is calculated. The nearest neighboring UAV monitoring node in terms of spatial distance is selected as the assignment target. The hovering point to be assigned is inserted into the next position after the current position in the hovering point sequence of the neighboring UAV monitoring node. Simultaneously, the task flight path of the neighboring UAV monitoring node is updated. The task flight path update uses an adaptive ant colony algorithm for replanning. During planning, the hovering points currently reached by the neighboring UAV monitoring node are considered as fixed visited nodes, and the updated hovering point sequence is considered as all nodes to be visited, thus regenerating the shortest closed flight path. For each drone monitoring node that triggers the condition, the above allocation is performed on all its unvisited hovering points until all hovering points to be allocated have been reallocated. Then, the original drone monitoring node directly plans the shortest return route from its current position back to the take-off and landing platform and executes the return. Optionally, if there are no adjacent drone monitoring nodes with sufficient remaining power around a hovering point to be allocated, the drone monitoring node abandons access to that hovering point, and the next round of the task cycle calls upon a supplementary monitoring node to complete the coverage.

[0097] In practical implementation, after solving the optimization function of the UAV monitoring network layout to obtain the optimal spatial coordinate set of the UAV monitoring nodes, spatial distribution data and activity intensity timetables of human activity disturbance sources in the mangrove monitoring area are obtained. The spatial distribution data of human activity disturbance sources comes from the functional zoning map and human activity monitoring report provided by the mangrove wetland ecosystem management department, including vector line data of fishing boat channels, vector surface data of aquaculture pond boundaries, vector line data of tourist trails, and vector point data of settlements. The activity intensity timetable is the hourly activity intensity index corresponding to each disturbance source. The activity intensity index is calculated in the following ways: for fishing boat channels, the number of vessels passing through each hour is counted based on historical data from the Automatic Identification System (AIS) and normalized to the range of 0 to 1; for aquaculture ponds, the activity intensity index is set to 1 during the operation period and 0 at other times according to the feeding and harvesting time windows recorded in the aquaculture operation log; for tourist trails, it is generated by normalization based on the opening time of the scenic area and the statistical data of tourist flow; for settlements, hourly activity factors are generated based on population density distribution and work and rest patterns.

[0098] Anthropogenic interference risk scores are calculated for each uniform spatial grid based on the spatial distribution data and activity intensity timeline of human-induced interference sources. For each type of interference source, its spatial distribution vector data is rasterized and transformed into a raster layer consistent with the uniform spatial grid. The type-based interference intensity value for each uniform spatial grid is assigned as an inverse distance weighted value between the uniform spatial grid and the nearest spatial distance to the interference source. The nearest spatial distance to the interference source's vector line or vector boundary is calculated. ,when Less than or equal to the maximum influence distance of the interference source At that time, the type interference intensity value of the unified spatial grid is ,when Greater than At that time, the type of interference intensity value is 0; the maximum influence distance of the interference source is... Setting up different channels based on the type of interference source: Fishing boat navigation channel It is 500 meters long, aquaculture pond. It is 300 meters long, a tourist trail. 200 meters, residential area The distance is 1000 meters. Then, the intensity values ​​of all types of interference in each uniform spatial grid are summed and truncated to no more than 1 to obtain the static anthropogenic interference risk baseline score. Combined with the activity intensity timeline, the... Human interference risk score for each monitoring period It is the weighted average of the static human interference risk baseline score and the activity intensity index of all interference sources during the current monitoring period. When the activity intensity timetable indicates that a certain interference source is in a high-intensity activity period, the corresponding type of interference intensity value remains unchanged, and when it is in a low-intensity activity period, it is multiplied by an attenuation factor of 0.2.

[0099] Score the risk of human interference for each uniform spatial grid. Ecological sensitivity score in spatiotemporal distribution map Perform weighted fusion to generate a fusion risk score. The weighted fusion calculation formula is as follows:

[0100]

[0101] in, For the integrated risk score, the value range is from 0 to 1; To obtain the ecological sensitivity score value of the corresponding unified spatial grid in the spatiotemporal distribution map of mangrove ecological sensitivity, the score value of the time slice corresponding to the current monitoring period is extracted from the spatiotemporal distribution map. Risk scoring based on human interference; The weighting coefficient for ecological sensitivity is set at 0.6. This value is based on the following: ecological sensitivity reflects the inherent vulnerability of mangrove wetland ecosystems, and its contribution to the risk score should be slightly higher than that of dynamically changing anthropogenic disturbance factors. Through expert scoring, the weighting of ecological sensitivity is determined to be 0.6, while the weighting of anthropogenic disturbance risk is 0.4. Risk score for merging all uniform spatial graticles Arranged according to spatial coordinates, a fusion risk score distribution map is generated.

[0102] Spatial clustering of UAV monitoring nodes in the optimal spatial coordinate set is performed based on the fused risk score distribution map. The density-based spatial clustering algorithm DBSCAN is used for clustering, and the neighborhood radius of the DBSCAN algorithm is... Set to 200 meters, minimum number of nodes in the neighborhood of the core point. Set to 3. Use the horizontal coordinates of the hovering points of all UAV monitoring nodes as the input point set, and set the weight of each input point to the fusion risk score of its corresponding unified spatial grid. After DBSCAN clustering, it identified clusters whose fusion risk scores exceeded a preset high-risk threshold. Risk hotspots The value is set to 0.75, based on the following: when the fusion risk score exceeds 0.75, it indicates that the area simultaneously possesses high ecological sensitivity and high risk of human interference, requiring increased monitoring resource allocation. For each identified risk hotspot area cluster, the geometric center point of the horizontal coordinates of all hovering points within the cluster is calculated. The coordinates of the geometric center point are the average of the x-coordinates and y-coordinates of all points within the cluster. An additional drone monitoring node is added at the geometric center point, and its elevation coordinates are assigned to the flight altitude layer of 80 meters. The coordinates of all the additional drone monitoring nodes are added to the optimal spatial coordinate set to obtain the augmented optimal spatial coordinate set.

[0103] See Figure 10 The figure shows the ecological sensitivity scores corresponding to multiple uniform spatial grids within the mangrove monitoring area in the form of a scatter plot. Integration risk score The relationship between the two axes. The horizontal axis represents the ecological sensitivity score. The value ranges from 0 to 1, and the vertical axis represents the fusion risk score. The value range is also from 0 to 1. Each dot in the figure represents the score correspondence of a unified spatial grid. The overall trend shows a positive correlation from the lower left to the upper right, indicating that the integration risk score generally increases with the increase of the ecological sensitivity score.

[0104] The figure includes a weighted baseline. The dashed line, which starts from the origin of the coordinate system. Extend to point This reflects the theoretical weighting line with an ecological sensitivity weight of 0.6 in the fusion score. Most scatter points are distributed above this weighting baseline, indicating that the weighted contribution of human interference risk score is generally positive, making the fusion risk score higher than the predicted value of the simple weighting line, reflecting the role of human interference in enhancing overall risk.

[0105] The high-risk threshold line is also marked on the map. The dashed line is located at 0.75 on the vertical axis. A small portion of the scattered points exceed this threshold line, indicating that these uniform spatial grids have a high fusion risk score during the current monitoring period. They represent areas with both high ecological sensitivity and high risk of human interference, and are considered risk hotspots requiring key monitoring.

[0106] Overall, the data in the figure illustrates the spatial distribution characteristics of the fusion risk scores, reflecting both the inherent distribution patterns of ecological sensitivity and the dynamic impact of human interference. This aligns with the technical solution in Example 5, which generates a fusion risk score distribution map based on a weighted fusion of ecological sensitivity scores and human interference risk scores. The central tendency of the score points and the key threshold intervals in the figure provide a clear basis for subsequent spatial clustering of UAV monitoring nodes and identification of risk hotspots based on the fusion risk scores.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for network layout of unmanned aerial vehicle monitoring based on mangrove ecological monitoring, characterized in that, The method includes: Obtain multi-source historical monitoring data of mangrove wetland ecosystems, and construct a spatiotemporal distribution map of mangrove ecological sensitivity based on the multi-source historical monitoring data; The gradient change boundary of ecological risk level in mangrove ecosystem is identified based on the spatiotemporal distribution map, and the initial deployment density of UAV monitoring network in different ecological risk level areas is determined based on the gradient change boundary. Meteorological and tidal forecast data of the mangrove monitoring area are collected, and the initial deployment density is dynamically adjusted in time according to the meteorological and tidal forecast data to generate a dynamic deployment density matrix for each monitoring period. Using the dynamic deployment density matrix as a constraint, a UAV monitoring network layout optimization function is constructed with the dual objectives of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. Solving the UAV monitoring network layout optimization function yields the optimal set of spatial coordinates for UAV monitoring nodes. Based on the optimal set of spatial coordinates, generate the mission flight path and hovering point sequence for each UAV monitoring node.

2. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 1, characterized in that, The acquisition of multi-source historical monitoring data of mangrove wetland ecosystems, and the construction of a spatiotemporal distribution map of mangrove ecological sensitivity based on the multi-source historical monitoring data, specifically includes: Historical satellite remote sensing image data, historical ground sensor monitoring data, and historical manual patrol data of mangrove wetland ecosystems are retrieved. The satellite remote sensing image data, the historical ground sensor monitoring data, and the historical manual patrol data are then processed by timestamp alignment and spatial coordinate normalization to obtain spatiotemporally aligned multi-source historical monitoring data. The historical change sequence of mangrove vegetation cover, the historical distribution range of benthic habitat, and the historical evolution trajectory of tidal channel system are extracted from the spatiotemporally aligned multi-source historical monitoring data. The historical change sequence of mangrove vegetation cover, the historical distribution range of benthic habitat, and the historical evolution trajectory of tidal channel system are input into a pre-trained ecological sensitivity assessment model, and the ecological sensitivity score of each spatial grid in the mangrove wetland ecosystem is output. The ecological sensitivity scores of all spatial grids are stitched together according to spatial coordinates, and the stitched ecological sensitivity scores are averaged along the time axis using a sliding window to generate a spatiotemporal distribution map of mangrove ecological sensitivity.

3. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 2, characterized in that, The ecological sensitivity assessment model is a spatiotemporal fusion model built on long short-term memory networks and convolutional neural networks.

4. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 2, characterized in that, Based on the spatiotemporal distribution map, the gradient change boundary of ecological risk level in the mangrove ecosystem is identified. Based on this gradient change boundary, the initial deployment density of the UAV monitoring network in different ecological risk level regions is determined, specifically including: The ecological sensitivity score values ​​in the spatiotemporal distribution map of the mangrove ecological sensitivity are classified at equal intervals according to their numerical values. Each spatial grid is assigned to the corresponding ecological risk level to obtain the ecological risk level distribution map of the mangrove wetland ecosystem. The ecological risk level differences of all adjacent spatial grids are traversed on the ecological risk level distribution map. When the ecological risk levels of two adjacent spatial grids are different, the common boundary of the two adjacent spatial grids is marked as the gradient change boundary. The area corresponding to each ecological risk level is statistically analyzed. Based on the product of the area and the preset benchmark monitoring density coefficient, the initial deployment density of the UAV monitoring network in each ecological risk level area is calculated. The benchmark monitoring density coefficient increases with the increase of the ecological risk level.

5. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 4, characterized in that, The process involves collecting meteorological and tidal forecast data for the mangrove monitoring area, dynamically adjusting the initial deployment density based on this data, and generating a dynamic deployment density matrix for each monitoring period. Specifically, this includes: Hourly wind speed forecast data, hourly rainfall probability forecast data, hourly high tide time data, and hourly tidal range data for the mangrove monitoring area during the future monitoring period were collected from the meteorological forecast interface and the tidal forecast interface, respectively. The hourly wind speed forecast data, the hourly rainfall probability forecast data, the hourly high tide time data, and the hourly tidal range data are fused according to the same timestamp to obtain the comprehensive environmental disturbance vector for each monitoring period. The initial deployment density of each ecological risk level area and the comprehensive environmental disturbance vector of the corresponding monitoring period are input into the preset environmental attenuation function to calculate the dynamic adjusted density of each ecological risk level area in each monitoring period. The dynamic adjusted densities of all ecological risk level areas are arranged according to spatial coordinates to generate the dynamic deployment density matrix for each monitoring period.

6. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 5, characterized in that, The environmental attenuation function is a multivariate nonlinear regression function with wind speed, rainfall probability, high tide time offset, and tidal range as input parameters.

7. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 5, characterized in that, The process involves constructing a UAV monitoring network layout optimization function with the dynamic deployment density matrix as a constraint, aiming to maximize monitoring coverage integrity and minimize monitoring energy consumption. Solving this optimization function yields the optimal set of spatial coordinates for the UAV monitoring nodes. Specifically, this includes: The mangrove monitoring area is discretized into a grid of candidate monitoring points, and the required probability weight of each candidate monitoring point is determined according to the dynamic deployment density matrix. The sub-objective function for maximizing monitoring coverage integrity is defined as the ratio of the union of the coverage areas of all selected UAV monitoring nodes to the total area of ​​the mangrove monitoring area, where the coverage area of ​​each UAV monitoring node is determined by its sensor detection radius and flight altitude. The sub-objective function for minimizing monitoring energy consumption is defined as the sum of the flight energy consumption of all selected UAV monitoring nodes from the take-off and landing platform to their hovering point and the hovering energy consumption of each node when performing monitoring tasks at the hovering point. The required probability weights are embedded as constraints into the sub-objective functions of maximizing monitoring coverage integrity and minimizing monitoring energy consumption. The multi-objective particle swarm optimization algorithm is used to solve the sub-objective functions of maximizing monitoring coverage integrity and minimizing monitoring energy consumption in parallel iteratively, and the optimal spatial coordinate set of the UAV monitoring node corresponding to the optimal solution on the Pareto front is output.

8. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 7, characterized in that, The multi-objective particle swarm optimization algorithm introduces an adaptive inertia weight factor and a mutation operator to escape local optima in the later stages of iteration.

9. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 7, characterized in that, The step of generating the mission flight path and hovering point sequence for each UAV monitoring node based on the optimal spatial coordinate set specifically includes: Extract the three-dimensional coordinates of the hovering point of each UAV monitoring node from the optimal set of spatial coordinates, and use the take-off and landing platform coordinates of each UAV monitoring node as the starting coordinates and ending coordinates of the flight path; For each UAV monitoring node, the starting point coordinates of the flight path are used as the starting point, the three-dimensional coordinates of all hovering points of the UAV monitoring node are used as intermediate points, and the ending point coordinates of the flight path are used as the ending point. The adaptive ant colony algorithm is used to plan the shortest closed flight path that passes through the three-dimensional coordinates of all hovering points and outputs the task flight path of each UAV monitoring node. Arrange the three-dimensional coordinates of all hovering points of each UAV monitoring node according to their access order on the mission route to generate a hovering point sequence for each UAV monitoring node.

10. The method for deploying a drone monitoring network based on mangrove ecological monitoring according to claim 9, characterized in that, The adaptive ant colony algorithm employs a dynamically updated pheromone evaporation coefficient, which adaptively decreases as the remaining power of the UAV monitoring node decreases.

Images