Mine ecological damage intelligent identification method and system based on remote sensing and support vector machine

Abstract

The invention provides a mine ecological damage intelligent identification method and system based on remote sensing and a support vector machine, and relates to the technical field of mine ecological environment monitoring. The method comprises the steps of collecting multi-source remote sensing data to construct an initial data set, extracting comprehensive ecological damage features through a three-dimensional residual convolutional neural network, and realizing nonlinear classification by using a mixed kernel function support vector machine; meanwhile, an ecological damage index is calculated, a distribution diagram is generated, damage grades are divided by combining a dynamic threshold value, and a three-level early warning mechanism is triggered to generate restoration suggestions. According to the method, by introducing a deep learning model and a multi-source remote sensing data fusion technology, the expression ability of ecological characteristics is enhanced; through a support vector machine algorithm and kernel function mapping, the adaptability of the classification model in a complex scene is improved; through a dynamic threshold value and a multi-stage early warning mechanism, accurate grading and timely response of ecological damage are realized, and a scientific basis and technical support are provided for monitoring and restoration of a mine ecological environment.

Description

Technical Field

[0001] The present invention relates to the technical field of mine ecological environment monitoring, and in particular to a method and system for intelligently identifying mine ecological damage based on remote sensing and support vector machines. Background Art

[0002] With the rapid development of remote sensing technology and machine learning algorithms, ecological damage identification methods based on remote sensing imagery and intelligent algorithms have gradually become a research hotspot in the field of mine ecological environment monitoring. Rapid identification and assessment of ecological damage during mining operations are crucial for achieving sustainable development of mining environments. However, existing methods for identifying ecological damage in mines still have many shortcomings in data processing, feature extraction, and classification model construction, which affect identification accuracy and application efficiency.

[0003] Currently, patent publication number CN113744249B discloses a method for investigating marine ecological damage. This method constructs a remote sensing survey information database for marine ecological damage, combines it with optimal remote sensing observation methods to acquire data, and utilizes image preprocessing and feature extraction techniques to generate damage survey results. However, this method primarily relies on traditional image processing methods for feature extraction of remote sensing images, failing to fully utilize the advantages of advanced classification algorithms such as support vector machines (SVMs). This results in insufficient accuracy in identifying ecological damage in complex scenarios. Therefore, it is crucial to design an intelligent method and system for identifying mine ecological damage based on remote sensing and support vector machines. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for intelligent identification of mine ecological damage based on remote sensing and support vector machine, so as to improve the adaptability of the classification model in complex scenarios.

[0005] To achieve the above object, the present invention provides the following solutions: A method for intelligently identifying mine ecological damage based on remote sensing and support vector machines includes the following steps: In the mining area and surrounding ecologically sensitive areas, multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery were simultaneously collected to obtain an initial dataset. The initial dataset was preprocessed, and ecological features including spatial, spectral, and texture features were extracted using a three-dimensional convolutional neural network. The ecological features were then fused with auxiliary topographic, geological, and meteorological data to obtain an ecological damage feature set. Based on the ecological damage feature set, a mine ecological damage classification model was constructed using the support vector machine algorithm. The nonlinear classification problem of the mine ecological damage classification model was solved through kernel function mapping to obtain the ecological damage type and spatial distribution. Calculate the ecological damage index of the mining area according to the ecological damage type, and combine it with the spatial analysis algorithm to obtain the ecological damage value, and then normalize the ecological damage value; The ecological damage level is divided according to the normalized ecological damage value. When the ecological damage value reaches or exceeds the preset threshold, the multi-level early warning mechanism is triggered and ecological restoration recommendations are generated.

[0006] Optionally, multi-source remote sensing data, including high-resolution satellite imagery, multispectral imagery, and radar imagery, can be collected simultaneously within the mining area and surrounding ecologically sensitive areas to obtain an initial dataset, including: According to the mining scope and terrain diffusion model, the data collection points in the mining core area, transition area and ecological protection area are determined respectively; Obtain high-resolution satellite images of data collection points through satellite remote sensing platforms and extract information on land cover changes; The multispectral images of the data collection points were collected by a hyperspectral instrument carried by an unmanned aerial vehicle, and the vegetation stress index, water index and building index were calculated; Extracting surface deformation and soil moisture from radar images; The land cover change information, vegetation stress index, water body index, building index, surface deformation and soil moisture were aligned in time and space to obtain the initial dataset.

[0007] Optionally, the initial data set is preprocessed and ecological features including spatial features, spectral features, and texture features are extracted using a three-dimensional convolutional neural network. The ecological features are then fused with auxiliary terrain, geological, and meteorological data to obtain an ecological damage feature set, including: Perform change vector analysis on the initial data set to extract the key phase nodes of the time series; Construct a three-dimensional convolutional neural network consisting of spatial convolution layer, spectral convolution layer, texture convolution layer and feature fusion layer; The spatial features including mining surface boundary vector, dump expansion area and surface settlement gradient are obtained based on the surface deformation gradient of key time nodes extracted by the spatial convolution layer. The spectral derivatives of the key time-phase nodes extracted by the spectral convolution layer are used to obtain the spectral features including the vegetation red edge position offset, mineral characteristic absorption depth and water turbidity index; The texture features of key time phase nodes, including surface fragmentation, fracture network density and vegetation distribution heterogeneity index, are extracted based on the texture convolution layer. The spatial features, spectral features and texture features are concatenated and compressed to obtain ecological features.

[0008] Optionally, perform change vector analysis on the initial dataset to extract key phase nodes of the time series, including: Perform difference operation on any two data of different time phases in the initial data set in the multi-dimensional spectral space to obtain the change vector; Calculate the amplitude and direction angle of the change vector; the amplitude is used to indicate the severity of the data change, and the direction angle is used to indicate the spectral type of the data change; Identify the nodes where the amplitude has a local peak, and select the nodes whose direction angle is the same as the preset direction angle as the key phase nodes.

[0009] Optionally, based on the ecological damage feature set, a mine ecological damage classification model is constructed using a support vector machine algorithm, and the nonlinear classification problem of the mine ecological damage classification model is solved through kernel function mapping to obtain the ecological damage type and spatial distribution, including: Generate a historical dataset based on historical image data of the mining area and surrounding ecologically sensitive areas during the non-mining period; The Gaussian radial basis kernel function was used as the kernel function of the support vector machine model, and the parameters of the support vector machine model were optimized by maximizing the objective function of the classification interval to obtain the ecological damage classification model. The correlation between the ecological damage feature set and the historical data set is calculated through the ecological damage classification model, and the ecological damage type and spatial distribution are obtained based on the correlation.

[0010] Optionally, the ecological damage index of the mining area is calculated according to the ecological damage type, and the ecological damage value is obtained by combining the spatial analysis algorithm, and the ecological damage value is normalized, including: According to the type of ecological damage, the ecological damage index of each spatial grid unit in the mining area is calculated separately; the calculation formula of the ecological damage index is: ,in, is the weight coefficient, is the probability of damage of type c; The ecological damage index was modified by spatial autocorrelation analysis and kernel density estimation to obtain the modified damage index. The ecological damage value is calculated based on the modified damage index and environmental sensitivity factor; the calculation formula for the ecological damage value is: ,in, To correct the damage index, is the sensitive area adjustment coefficient, For sensitive area level, is the rainfall erosivity of the current grid, is the regional benchmark erosivity.

[0011] The ecological damage value was normalized using the maximum-minimum normalization method.

[0012] Optionally, the ecological damage index is corrected by spatial autocorrelation analysis and kernel density estimation to obtain a corrected damage index, including: The spatial autocorrelation index is calculated based on the ecological damage index; the calculation formula of the spatial autocorrelation index is: ,in, is the spatial weight matrix, n is the number of neighborhood grids, is the average value of the ecological damage index; The damage diffusion density of the spatial autocorrelation index is obtained through the Gaussian kernel function; The modified damage index is obtained based on the damage diffusion density and spatial autocorrelation index; the calculation formula of the modified damage index is: ,in To damage the diffusion density.

[0013] Optionally, the ecological damage level is divided according to the normalized ecological damage value. When the ecological damage value reaches or exceeds a preset threshold, a multi-level early warning mechanism is triggered and ecological restoration suggestions are generated, including: When the ecological damage value is within the range of [0,3), the ecological damage level is judged as mild damage; when the ecological damage value is within the range of [3,6), the ecological damage level is judged as moderate damage; when the ecological damage value is within the range of [6,8), the ecological damage level is judged as severe damage; when the ecological damage value is greater than or equal to 8, the ecological damage level is judged as extremely severe damage; The dynamic threshold is generated based on environmental parameters, including rainfall erosivity index and terrain relief. The dynamic threshold is calculated as follows: , where TI is the terrain relief and RI is the rainfall erosivity index; When the ecological damage value ≥ A blue warning is generated when the ecological damage value ≥ max( , 3.5), a yellow warning is generated; when the ecological damage value is ≥6, an orange warning is generated; when the ecological damage value is ≥8 or When it is greater than 0.5, a red alert is generated.

[0014] An intelligent identification system for mine ecological damage based on remote sensing and support vector machine, including: The data acquisition module is used to simultaneously collect multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery within the mining area and surrounding ecologically sensitive areas to obtain an initial data set; The data preprocessing module is used to preprocess the initial data set, extract ecological features including spatial features, spectral features, and texture features through a three-dimensional convolutional neural network, and fuse the ecological features with topographic, geological, and meteorological auxiliary data to obtain an ecological damage feature set; The classification modeling module is used to construct a mine ecological damage classification model based on the ecological damage feature set through the support vector machine algorithm, and solve the nonlinear classification problem of the mine ecological damage classification model through kernel function mapping to obtain the ecological damage type and spatial distribution; The index calculation module is used to calculate the ecological damage index of the mining area according to the ecological damage type, and obtain the ecological damage value by combining the spatial analysis algorithm, and normalize the ecological damage value; The early warning generation module is used to classify ecological damage levels according to the normalized ecological damage value. When the ecological damage value reaches or exceeds the preset threshold, the multi-level early warning mechanism is triggered and ecological restoration suggestions are generated.

[0015] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the present invention provides an intelligent identification method for mine ecological damage based on remote sensing and support vector machines, the method comprising: synchronously collecting multi-source remote sensing data including high-resolution satellite images, multispectral images and radar images in the mining area and the surrounding ecologically sensitive areas to obtain an initial data set; preprocessing the initial data set, extracting ecological features including spatial features, spectral features and texture features through a three-dimensional convolutional neural network, and fusing the ecological features with terrain, geological and meteorological auxiliary data to obtain an ecological damage feature set; based on the ecological damage feature set, constructing a mine ecological damage classification model through a support vector machine algorithm, and solving the nonlinear classification problem of the mine ecological damage classification model through kernel function mapping to obtain the ecological damage type and spatial distribution; calculating the ecological damage index of the mine area according to the ecological damage type, and obtaining the ecological damage value by combining the spatial analysis algorithm, and normalizing the ecological damage value; dividing the ecological damage level according to the normalized ecological damage value, and triggering a multi-level early warning mechanism and generating ecological restoration suggestions when the ecological damage value reaches or exceeds a preset threshold. This method enhances the expression ability of ecological characteristics by introducing deep learning models and multi-source remote sensing data fusion technology; improves the adaptability of classification models in complex scenarios through support vector machine algorithms and kernel function mapping; and achieves accurate classification and timely response to ecological damage through dynamic thresholds and multi-level early warning mechanisms, providing a scientific basis and technical support for mine ecological environment monitoring and restoration. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1This is a flow chart of the method for intelligently identifying mine ecological damage according to the present invention; Figure 2 A flow chart is constructed for the ecological damage feature set of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown, the present invention provides an intelligent identification method for mine ecological damage based on remote sensing and support vector machine, comprising the following steps: Step 100: synchronously collect multi-source remote sensing data including high-resolution satellite images, multispectral images, and radar images within the mining area and surrounding ecologically sensitive areas to obtain an initial data set; Specifically, this embodiment first calculates the impact range based on the mining center coordinates and the digital elevation model (DEM). With the mining point as the center, a circular area with a radius of 0.5 times the mining license area is used as the core mining area; then, according to the terrain diffusion equation, Determine the boundaries of the transition zone, where K is the diffusion coefficient, which is 1.2 in this embodiment, α is the terrain attenuation factor, which is 0.5 in this embodiment. The elevation gradient was calculated; the ecological red line boundary was expanded 200 meters to form an ecological protection zone. After the division, data collection points were arranged in a 1km×1km grid within the three areas, ensuring that each grid contained at least one collection point.

[0021] High-resolution satellite images of the data collection points were then acquired through a satellite remote sensing platform. The image was segmented using the object-oriented classification method of the eCognition platform, with a segmentation scale parameter of 50, a shape factor of 0.3, and a compactness of 0.7. The Normalized Difference Vegetation Index (NDVI) was then calculated for the images of the two time phases using the following formula: , where NIR is the near-infrared band and Red is the red band. Areas with a NDVI difference greater than 0.2 between the two time phases are marked as areas of land cover change, and the NDVI difference reflects land cover change information. Multispectral images of data collection points are collected using an unmanned aerial vehicle (UAV) hyperspectral instrument, and the vegetation stress index, water index, and building index are calculated. The vegetation stress index is the sum of the red edge position offset (REIP) and the chlorophyll fluorescence index (SIF), and the calculation formulas are: , ,in is the reflectivity at wavelength λ, It is the reference value of the undisturbed area. The calculation formula of water body index TSI is: , where Green is the green band. The calculation formula of the building index is: ,in, is the pixel value in the gray-level co-occurrence matrix (GLCM) i The probability of a land cover change is 0. The surface deformation and soil moisture of the same area in the radar image are also extracted. Finally, the coordinate system of all images is converted to the UTM projection using bilinear interpolation to ensure the alignment of the grid center coordinates. Six types of data—land cover change information, vegetation stress index, water index, building index, surface deformation, and soil moisture—are aggregated into a multidimensional matrix by grid, resulting in the initial dataset.

[0022] Step 200: Preprocess the initial data set, extract ecological features including spatial features, spectral features and texture features through a three-dimensional convolutional neural network, and fuse the ecological features with terrain, geological and meteorological auxiliary data to obtain an ecological damage feature set; the specific steps are as follows: Figure 2 As shown, including: Step 201: Perform change vector analysis on the initial data set to extract key phase nodes of the time series; Specifically, the method uses a difference operation between any two data points at different times in the initial dataset in a multidimensional spectral space to generate a change vector. The magnitude and azimuth of this change vector are then calculated to quantify the severity of the data change and its spectral characteristics. The amplitude and azimuth are then screened, and the time points where the amplitude reaches a local peak and the azimuth matches the preset azimuth are identified as key time points.

[0023] Step 202: constructing a three-dimensional convolutional neural network including a spatial convolution layer, a spectral convolution layer, a texture convolution layer, and a feature fusion layer; Specifically, the convolution kernel sizes of the spatial convolution layer, spectral convolution layer, and texture convolution layer are 3×3×3, 1×1×X, and 5×5×1, respectively.

[0024] Step 203: obtaining spatial features including mining face boundary vector, dump expansion area, and surface settlement gradient based on the mining area surface deformation gradient of the key time phase nodes extracted by the spatial convolution layer; Specifically, the first convolution layer of the spatial convolution layer uses 32 3×3×3 convolution kernels, and its activation function is ReLU; the second convolution layer uses 64 3×3×3 convolution kernels to extract the surface deformation gradient characteristics of the mining area, which is expressed as: ,in is the initial data set, G It is a spatial feature. In this embodiment, the spatial features include: mining surface boundary vector, spoil dump expansion area and surface settlement gradient.

[0025] Step 204: obtaining spectral features including vegetation red edge position offset, mineral characteristic absorption depth, and water turbidity index based on the spectral derivatives of the key time phase nodes extracted by the spectral convolution layer; Specifically, the first convolution layer of the spectral convolution layer uses a 1×1×8 convolution kernel to compress the spectral dimension of the initial data set to 1 / 4 of the original band; the second convolution layer uses a 1×1×3 convolution kernel to extract the spectral derivative. , the expression is: ,in and is the influence data of two adjacent frames in the initial data set, is the wavelength difference between two adjacent frames. Then the mineral characteristic absorption depth is obtained according to the spectral derivative , the calculation formula is: The red edge position offset, mineral characteristic absorption depth and water turbidity index are aggregated into spectral features. The water turbidity index is the ratio of the wavelength sum of the 450-650nm band to the wavelength sum of the 650-900nm band.

[0026] Step 205: extracting texture features including surface fragmentation, fracture network density, and vegetation distribution heterogeneity index from key time phase nodes based on the texture convolution layer; Specifically, the primary texture of the initial dataset is first extracted through a 5×5×1 convolution kernel to output a feature map. The multi-directional texture response of the feature map is then enhanced through a Gabor filter to obtain texture features including surface fragmentation, crack network density, and vegetation distribution heterogeneity index.

[0027] Furthermore, the high-resolution satellite image is grayscaled and divided into grid cells of fixed size. Then, the gray-level co-occurrence matrix is ​​calculated in each window, and the joint probability distribution of pixel gray values ​​in specific directions and distances is statistically analyzed. Then, the texture entropy value is extracted based on the GLCM matrix. The texture entropy value calculation formula is: ,in Indicates grayscalei and j The co-occurrence probability is finally calculated by the formula Map the entropy value to the [0,1] interval to obtain the surface fragmentation The closer the value is to 1, the greater the degree of surface patch fragmentation due to mining activities. The fracture network density is the ratio of the total length of surface fractures to the grid area. The vegetation distribution heterogeneity index is the total area of ​​all vegetation patches within a grid cell.

[0028] Step 206: perform feature concatenation and dimensionality reduction compression on the spatial features, spectral features, and texture features to obtain ecological features.

[0029] Specifically, the spatial features, spectral features, and texture features are spliced ​​according to the channel dimension to generate a fused feature map, which is then compressed to 128 dimensions through a 1×1×1 convolution kernel, and the ecological characteristics are obtained through a Sigmoid activation function and a 3D convolution operation.

[0030] Step 207: Fuse the ecological characteristics with the terrain, geology, and meteorological auxiliary data to obtain an ecological damage feature set.

[0031] Step 300: Based on the ecological damage feature set, a mine ecological damage classification model is constructed using a support vector machine algorithm, and the nonlinear classification problem of the mine ecological damage classification model is solved through kernel function mapping to obtain the ecological damage type and spatial distribution; Specifically, a historical dataset is first generated based on historical image data from the mining area and surrounding ecologically sensitive areas during the pre-mining period. A Gaussian radial basis kernel function is then used as the kernel function of a support vector machine model. The support vector machine model parameters are optimized using the objective function of maximizing the classification interval to obtain an ecological damage classification model. Finally, the ecological damage classification model calculates the correlation between the ecological damage feature set and the historical dataset, and based on this correlation, the ecological damage type and spatial distribution are determined. In this embodiment, the ecological damage types include vegetation destruction, water and soil pollution, surface subsidence, and biodiversity loss.

[0032] Furthermore, the expression of the Gaussian radial basis kernel function is: ,in x i and x j Represents the feature vector of the dataset. The expression of the objective function is: ,in is the normal vector of the classification hyperplane in the support vector machine model, b is the bias term. In this embodiment, the correlation between the ecological damage feature set and the historical dataset is represented by the Euclidean distance between the two sets. In this embodiment, the parameter optimization process uses a particle swarm optimization algorithm for iterative updates.

[0033] Step 400: Calculate the ecological damage index of the mining area according to the ecological damage type, obtain the ecological damage value by combining the spatial analysis algorithm, and normalize the ecological damage value; Specifically, according to the type of ecological damage, the ecological damage index of each spatial grid unit in the mining area is calculated separately. The calculation formula is: ; in, is the weight coefficient, is the probability of damage of type c.

[0034] The ecological damage index is then corrected through spatial autocorrelation analysis and kernel density estimation to obtain the corrected damage index. Specifically, the spatial autocorrelation index is first calculated based on the ecological damage index, and the calculation formula is: ; in, is the spatial weight matrix, n is the number of neighborhood grids, is the average value of the ecological damage index; then the damage diffusion density of the spatial autocorrelation index is obtained through the Gaussian kernel function, and the expression is: ; Where h is the bandwidth, Representation Grid k With Grid i The Euclidean distance of .

[0035] Then, the modified damage index is obtained according to the damage diffusion density and spatial autocorrelation index. The calculation formula is: ; in is the damage diffusion density. Finally, the ecological damage value is calculated based on the modified damage index and environmental sensitivity factor. The calculation formula is: ; in, To correct the damage index, is the sensitive area adjustment coefficient, For sensitive area level, is the rainfall erosivity of the current grid, is the regional benchmark erosion power. The ecological damage value is normalized using the maximum and minimum normalization method.

[0036] Step 500: The ecological damage level is divided according to the normalized ecological damage value. When the ecological damage value reaches or exceeds the preset threshold, a multi-level early warning mechanism is triggered and an ecological restoration suggestion is generated.

[0037] Specifically, when the ecological damage value is within the range of [0,3), the ecological damage level is judged as mild damage; when the ecological damage value is within the range of [3,6), the ecological damage level is judged as moderate damage; when the ecological damage value is within the range of [6,8), the ecological damage level is judged as severe damage; when the ecological damage value is greater than or equal to 8, the ecological damage level is judged as extremely severe damage. Dynamic thresholds are generated based on environmental parameters, including rainfall erosion index and terrain relief. The calculation formula for the dynamic threshold is: ; Among them, TI is the terrain relief, RI is the rainfall erosion index. A blue warning is generated when the ecological damage value ≥ max( , 3.5), a yellow warning is generated; when the ecological damage value is ≥6, an orange warning is generated; when the ecological damage value is ≥8 or When it is greater than 0.5, a red alert is generated.

[0038] The present invention also provides an intelligent identification system for mine ecological damage based on remote sensing and support vector machine, comprising: The data acquisition module is used to simultaneously collect multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery within the mining area and surrounding ecologically sensitive areas to obtain an initial data set; The data preprocessing module is used to preprocess the initial data set, extract ecological features including spatial features, spectral features, and texture features through a three-dimensional convolutional neural network, and fuse the ecological features with topographic, geological, and meteorological auxiliary data to obtain an ecological damage feature set; The classification modeling module is used to construct a mine ecological damage classification model based on the ecological damage feature set through the support vector machine algorithm, and solve the nonlinear classification problem of the mine ecological damage classification model through kernel function mapping to obtain the ecological damage type and spatial distribution; The index calculation module is used to calculate the ecological damage index of the mining area according to the ecological damage type, and obtain the ecological damage value by combining the spatial analysis algorithm, and normalize the ecological damage value; The early warning generation module is used to classify ecological damage levels according to the normalized ecological damage value. When the ecological damage value reaches or exceeds the preset threshold, the multi-level early warning mechanism is triggered and ecological restoration suggestions are generated.

[0039] The beneficial effects of the present invention are as follows: 1) By simultaneously collecting high-resolution satellite, multispectral, and radar images, combined with auxiliary topographic, geological, and meteorological data, this approach overcomes the feature loss problem caused by traditional methods relying on single remote sensing data, significantly improving the comprehensiveness and accuracy of ecological damage identification. 2) Using a 3D convolutional neural network to automatically extract spatial, spectral, and texture features, it overcomes the limitations of manually designed features and significantly improves the ability to express features in complex scenarios. 3) The support vector machine model was introduced and combined with the Gaussian radial basis kernel function. Through kernel mapping, the nonlinear classification problem of the mining environment was solved, and the classification model's recognition accuracy for complex damage types such as vegetation destruction and water and soil pollution was improved; 4) Innovatively designed ecological damage index (EDI) and ecological damage value (EDV) algorithms were used. The spatial clustering of damage was quantified by introducing spatial autocorrelation analysis, and the damage diffusion effect was characterized by kernel density estimation, making the assessment results more consistent with the spatiotemporal propagation patterns of ecological damage in mining areas.

[0040] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0041] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for intelligent identification of mine ecological damage based on remote sensing and support vector machine, characterized in that: The steps include: In the mining area and surrounding ecologically sensitive areas, multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery were simultaneously collected to obtain an initial dataset. Preprocessing the initial data set, extracting ecological features including spatial features, spectral features, and texture features through a three-dimensional convolutional neural network, and fusing the ecological features with auxiliary topographic, geological, and meteorological data to obtain an ecological damage feature set; Based on the ecological damage feature set, a mine ecological damage classification model is constructed using a support vector machine algorithm, and the nonlinear classification problem of the mine ecological damage classification model is solved through kernel function mapping to obtain the ecological damage type and spatial distribution; Calculating an ecological damage index for the mining area based on the ecological damage type, obtaining an ecological damage value by combining a spatial analysis algorithm, and normalizing the ecological damage value; The ecological damage level is divided according to the normalized ecological damage value. When the ecological damage value reaches or exceeds a preset threshold, a multi-level early warning mechanism is triggered and an ecological restoration suggestion is generated.

2. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 1 is characterized in that: In the mining area and surrounding ecologically sensitive areas, multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery were collected simultaneously to obtain an initial dataset, including: According to the mining scope and terrain diffusion model, the data collection points in the mining core area, transition area and ecological protection area are determined respectively; Acquiring the high-resolution satellite images of the data collection points through a satellite remote sensing platform and extracting land cover change information; Collecting multispectral images of the data collection points using a hyperspectral instrument carried by an unmanned aerial vehicle, and calculating a vegetation stress index, a water body index, and a building index; extracting ground surface deformation and soil moisture based on the radar image; The surface cover change information, the vegetation stress index, the water body index, the building index, the surface deformation and the soil moisture are aligned in time and space to obtain the initial data set.

3. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 1 is characterized in that: The initial data set is preprocessed, and ecological features including spatial features, spectral features, and texture features are extracted using a three-dimensional convolutional neural network. The ecological features are then fused with auxiliary terrain, geological, and meteorological data to obtain an ecological damage feature set, including: Performing change vector analysis on the initial data set to extract key phase nodes of the time series; Constructing the three-dimensional convolutional neural network comprising a spatial convolution layer, a spectral convolution layer, a texture convolution layer and a feature fusion layer; Obtaining the spatial features including the mining face boundary vector, the dump expansion area, and the surface settlement gradient according to the mining area surface deformation gradient of the key time phase node extracted by the spatial convolution layer; Obtaining the spectral features including vegetation red edge position offset, mineral characteristic absorption depth and water turbidity index according to the spectral derivatives of the key time phase nodes extracted by the spectral convolution layer; Extracting the texture features including surface fragmentation, fracture network density and vegetation distribution heterogeneity index in the key time phase nodes according to the texture convolution layer; The spatial features, the spectral features and the texture features are subjected to feature splicing and dimensionality reduction compression to obtain the ecological features.

4. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 3 is characterized in that: Perform change vector analysis on the initial data set to extract key phase nodes of the time series, including: Performing a difference operation on any two data of different time phases in the initial data set in a multi-dimensional spectral space to obtain a change vector; Calculating the amplitude and direction angle of the change vector; the amplitude is used to indicate the severity of the data change, and the direction angle is used to indicate the spectral type of the data change; The nodes where the amplitude has a local peak are identified, and the nodes whose direction angle is the same as the preset direction angle are selected as the key time phase nodes.

5. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 1 is characterized in that: Based on the ecological damage feature set, a mine ecological damage classification model is constructed using the support vector machine algorithm. The nonlinear classification problem of the mine ecological damage classification model is solved by kernel function mapping to obtain the ecological damage type and spatial distribution, including: Generate a historical data set based on historical image data of the mining area and the surrounding ecologically sensitive area during the non-mining period; Using a Gaussian radial basis kernel function as the kernel function of a support vector machine model, and optimizing the parameters of the support vector machine model by maximizing the objective function of the classification interval to obtain the ecological damage classification model; The correlation between the ecological damage feature set and the historical data set is calculated through the ecological damage classification model, and the ecological damage type and the spatial distribution are obtained based on the correlation.

6. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 1 is characterized in that: Calculating the ecological damage index of the mining area according to the ecological damage type, and combining the spatial analysis algorithm to obtain the ecological damage value, and normalizing the ecological damage value, including: According to the ecological damage type, the ecological damage index of each spatial grid unit in the mining area is calculated respectively; the calculation formula of the ecological damage index is: ,in, is the weight coefficient, is the probability of damage of type c; The ecological damage index is modified by spatial autocorrelation analysis and kernel density estimation to obtain a modified damage index; The ecological damage value is calculated based on the modified damage index and the environmental sensitivity factor; the calculation formula of the ecological damage value is: ,in, To correct the damage index, is the sensitive area adjustment coefficient, For sensitive area level, is the rainfall erosivity of the current grid, is the regional benchmark erosivity; The ecological damage value is normalized by the maximum and minimum normalization method.

7. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 6 is characterized in that: The ecological damage index is modified by spatial autocorrelation analysis and kernel density estimation to obtain a modified damage index, including: The spatial autocorrelation index is calculated based on the ecological damage index; the calculation formula of the spatial autocorrelation index is: ,in, is the spatial weight matrix, n is the number of neighborhood grids, is the average value of the ecological damage index; Obtaining the damage diffusion density of the spatial autocorrelation index through a Gaussian kernel function; The modified damage index is obtained according to the damage diffusion density and the spatial autocorrelation index; the calculation formula of the modified damage index is: ,in To damage the diffusion density.

8. The method for intelligent identification of mine ecological damage based on remote sensing and support vector machine according to claim 1 is characterized in that: The ecological damage level is divided according to the normalized ecological damage value. When the ecological damage value reaches or exceeds a preset threshold, a multi-level early warning mechanism is triggered and ecological restoration suggestions are generated, including: When the ecological damage value is within the range of [0,3), the ecological damage level is determined to be mild damage; when the ecological damage value is within the range of [3,6), the ecological damage level is determined to be moderate damage; when the ecological damage value is within the range of [6,8), the ecological damage level is determined to be severe damage; when the ecological damage value is greater than or equal to 8, the ecological damage level is determined to be extremely severe damage; A dynamic threshold is generated based on environmental parameters, including rainfall erosivity index and terrain relief. The dynamic threshold is calculated as follows: , where TI is the terrain relief and RI is the rainfall erosivity index; When the ecological damage value ≥ A blue warning is generated when the ecological damage value ≥ max( , 3.5), a yellow warning is generated; when the ecological damage value is ≥6, an orange warning is generated; when the ecological damage value is ≥8 or When it is greater than 0.5, a red alert is generated.

9. An intelligent identification system for mine ecological damage based on remote sensing and support vector machine, characterized in that: include: The data acquisition module is used to simultaneously collect multi-source remote sensing data including high-resolution satellite imagery, multispectral imagery, and radar imagery within the mining area and surrounding ecologically sensitive areas to obtain an initial data set; a data preprocessing module for preprocessing the initial data set, extracting ecological features including spatial features, spectral features, and texture features through a three-dimensional convolutional neural network, and fusing the ecological features with topographic, geological, and meteorological auxiliary data to obtain an ecological damage feature set; A classification modeling module is used to construct a mine ecological damage classification model based on the ecological damage feature set using a support vector machine algorithm, and solve the nonlinear classification problem of the mine ecological damage classification model through kernel function mapping to obtain the ecological damage type and spatial distribution; An index calculation module is used to calculate the ecological damage index of the mining area according to the ecological damage type, obtain the ecological damage value in combination with the spatial analysis algorithm, and normalize the ecological damage value; The early warning generation module is used to classify the ecological damage level according to the normalized ecological damage value, and when the ecological damage value reaches or exceeds a preset threshold, a multi-level early warning mechanism is triggered and an ecological restoration suggestion is generated.

Images