Mining area vegetation degradation identification method based on remote sensing spectrum

Abstract

The invention discloses a mining area vegetation degradation identification method based on a remote sensing spectrum. The method comprises the following steps: collecting multispectral image data of a mining area; based on the multispectral image data of the mining area, main spectral feature vectors are extracted by adopting a principal component analysis method, and a vegetation spectral feature set is determined; based on the vegetation spectral feature set, performing health state division on vegetation pixels by adopting a support vector machine classifier to obtain a vegetation degradation classification map; extracting a spatial distribution pattern from the vegetation degradation classification map, calculating an area proportion and a position coordinate of each type of pixels, and determining a spatial distribution feature of a degradation degree; acquiring a multi-temporal remote sensing image sequence according to the spatial distribution characteristics of the degradation degree, analyzing and tracking the change of a classification graph through time sequence difference, and judging a dynamic change trend; driving mechanism indexes are integrated from the dynamic change trend, and a comprehensive monitoring and recognition report is obtained and used for revealing the trend of vegetation degradation in the mining area.

Description

Technical Field

[0001] This invention belongs to the field of vegetation identification technology, and in particular relates to a method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy. Background Technology

[0002] Against the backdrop of prioritizing both ecological protection and resource development, monitoring vegetation degradation in mining areas has become an important research direction in environmental science. The impact of mining activities on the surrounding ecological environment cannot be ignored, especially since the health of vegetation directly relates to the stability and sustainable development of the regional ecosystem. Accurately assessing the degree of vegetation degradation in mining areas and revealing its changing trends is crucial for formulating effective ecological restoration strategies. Research in this field is not only an urgent need for environmental protection but also an important support for balancing resource development and ecological protection.

[0003] However, current research methods often fall short when faced with the complex environments of mining areas. Many methods struggle to comprehensively capture the interactive effects of multiple factors in vegetation degradation, particularly in integrating the dynamic changes in environmental disturbances and the characteristics of the vegetation itself. Especially for special areas like mining areas, relying solely on traditional ground surveys or single data sources makes it difficult to achieve high-precision dynamic monitoring over a wide area, and also fails to deeply reveal the underlying driving mechanisms of degradation.

[0004] A deeper technical challenge lies in how to effectively extract and utilize the spectral information of vegetation to reflect its health status. Spectral information, as an important indicator of vegetation health, can reveal the physiological state of vegetation through the reflectance characteristics of different wavelengths; however, its complexity and noise interference make accurate extraction exceptionally difficult. As research progresses, the processing of spectral information needs to be combined with the unique disturbance environment of mining areas, but this combination is often limited by the lack of systematic quantification of disturbance factors. For example, in mining areas, soil pollution and surface degradation can interfere with the spectral signals of vegetation. If the effects of these disturbances cannot be accurately distinguished, it is difficult to determine the true extent and specific causes of vegetation degradation.

[0005] Therefore, how to construct a monitoring system capable of accurately identifying the degree of vegetation degradation and revealing its spatial distribution and dynamic changes in complex mining environments, based on the refined processing of spectral information and comprehensively considering the multiple disturbances brought about by mining activities, has become a key issue that urgently needs to be addressed. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method for identifying vegetation degradation in mining areas based on remote sensing spectra. This method can accurately identify the degree of vegetation degradation and reveal its spatial distribution and dynamic changes within a monitoring system.

[0007] To achieve the above objectives, this invention provides a method for identifying vegetation degradation in mining areas based on remote sensing spectra, comprising:

[0008] Collect multispectral image data of the mining area;

[0009] Based on the multispectral image data of the mining area, principal component analysis was used to extract the main spectral feature vectors and determine the vegetation spectral feature set.

[0010] Based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels to obtain a vegetation degradation classification map.

[0011] The spatial distribution pattern is extracted from the vegetation degradation classification map, the area ratio and location coordinates of pixels in each category are calculated, and the spatial distribution characteristics of the degradation degree are determined.

[0012] Based on the spatial distribution characteristics of the degradation degree, multi-temporal remote sensing image sequences are acquired, and changes in the classification map are tracked through temporal difference analysis to determine the dynamic change trend;

[0013] By integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report is obtained to reveal the trend of vegetation degradation in mining areas.

[0014] Optionally, acquiring multispectral image data of the mining area includes:

[0015] Preliminary multispectral image data of the mining area were collected using a satellite remote sensing platform;

[0016] Using preset band filtering rules, specific bands related to vegetation cover are extracted from the preliminary multispectral image data to obtain the filtered band dataset.

[0017] Based on the filtered band dataset, an initial correction operation is performed to eliminate atmospheric interference and sensor bias, generating multispectral image data of the mining area.

[0018] Optionally, based on the multispectral image data of the mining area, principal component analysis is used to extract the main spectral feature vectors, and the vegetation spectral feature set is determined to include:

[0019] The multispectral image data of the mining area is standardized through data preprocessing to obtain a unified image data matrix;

[0020] Based on the unified image data matrix, principal component analysis is used to reduce the dimensionality of the spectral data, extract key spectral feature vectors, and determine the vegetation spectral feature set.

[0021] Optionally, based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels to obtain a vegetation degradation classification map, including:

[0022] If the reflectance of the feature vectors in the vegetation spectral feature set is lower than a preset threshold, the judgment result is obtained;

[0023] Based on the judgment results, the potential interference area is obtained;

[0024] The noise caused by soil pollution and surface damage in the potential interference area is removed by a mask filtering algorithm to obtain a purified spectral feature set.

[0025] Based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, thus obtaining the vegetation degradation classification map.

[0026] Optionally, noise caused by soil pollution and surface damage in the potential interference area is removed using a masking filtering algorithm to obtain a purified spectral feature set, including:

[0027] Based on the potential interference regions, obtain the interference region dataset;

[0028] The interference region dataset was processed using a mask filtering algorithm to separate the noise components caused by soil pollution and surface damage, resulting in a preliminarily purified spectral dataset.

[0029] Based on the preliminarily purified spectral dataset, the remaining abnormal reflectance points are detected. If abnormal points are detected, they are repaired by interpolation to obtain the repaired spectral dataset.

[0030] By smoothing the repaired spectral dataset to eliminate minor fluctuation noise, the purified spectral feature set is obtained.

[0031] Optionally, based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, to obtain the vegetation degradation classification map, including:

[0032] The purified spectral feature set is standardized according to a preset format to obtain a structured spectral feature dataset.

[0033] The support vector machine classifier is trained using the structured spectral feature dataset to obtain a classification model;

[0034] The health status of the vegetation pixels is classified using the classification model to obtain complete health status classification data.

[0035] Based on the complete health status classification data, vegetation pixel degradation distribution information is generated. The classification data is mapped to geographical location information using spatial mapping technology, and the vegetation degradation classification map is output.

[0036] Optionally, the spatial distribution pattern can be extracted from the vegetation degradation classification map, and the area ratio and location coordinates of pixels in each category can be calculated to determine the spatial distribution features of the degradation degree, including:

[0037] Based on the vegetation degradation classification map, the distribution information of pixels in each category is obtained by scanning pixels one by one;

[0038] Based on the distribution information of pixels of each category, the pixel area and area ratio of each type of degradation are calculated, and the statistical results of each type of pixel are determined by the grid division method.

[0039] Based on the area proportion data in the statistical results, the position coordinates of pixels of each category are obtained, and the clustering of each category in the spatial distribution is determined by spatial mapping technology.

[0040] Based on the aggregation in the spatial distribution, the spatial pattern of vegetation degradation is analyzed. If the pixel aggregation degree of a certain type of degradation is higher than a preset threshold, it is marked as a high-density degradation area, and the spatial distribution characteristics are obtained.

[0041] Based on the location coordinates and distribution characteristics of high-density degraded areas, and combined with the original data of the classified images, the random forest algorithm is used to further verify the spatial pattern and determine the spatial distribution law of the degree of degradation.

[0042] Based on the spatial distribution pattern, distribution characteristic data of degradation degree for each category are generated. Through data overlay processing, the spatial distribution characteristics of degradation degree are determined.

[0043] Optionally, time-series difference analysis can be used to track changes in the classification map and determine dynamic trends, including:

[0044] The temporal difference method is used to perform a layer-by-layer comparative analysis on the multi-temporal remote sensing image sequence to calculate the pixel-level differences between images and determine the range of the change area.

[0045] Based on the range of the changed area, classified image data is obtained, and the region is divided in combination with the spatial distribution characteristics to determine the evolution of the degree of degradation in each region.

[0046] Based on the described evolution, the dynamic trend of change is obtained.

[0047] Optionally, by integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report can be obtained, including:

[0048] By mapping the monitoring data in the dynamic trend, a comprehensive assessment result of vegetation degradation in the mining area is obtained;

[0049] Based on the comprehensive evaluation results and the analysis of the driving mechanism, analytical data on the mechanism of vegetation degradation in mining areas are generated.

[0050] The data is analyzed according to the mechanism to obtain the comprehensive monitoring and identification report.

[0051] Compared with the prior art, the present invention has the following advantages and technical effects:

[0052] This invention first acquires multispectral image data of the mining area using a satellite remote sensing platform, performs band selection and preliminary correction to obtain corrected images. Then, principal component analysis is used to extract the main spectral feature vectors, separating reflectance components related to vegetation health. If the reflectance of the feature vector is below a threshold, it is identified as a potential interference area, and a mask filtering algorithm is used to remove soil pollution and surface damage noise, obtaining a purified feature set. Next, a support vector machine classifier is used to classify the health status of vegetation pixels, generating a degradation classification map. Spatial distribution patterns are extracted from this map, and area proportions and location coordinates are calculated. Furthermore, multi-temporal image sequences are acquired, and changes are tracked through temporal difference analysis to determine dynamic trends. Finally, driving mechanism indicators such as mining intensity are integrated to generate a comprehensive monitoring report. This invention improves the accuracy and efficiency of monitoring vegetation degradation in mining areas by integrating principal component analysis, support vector machine, and temporal difference techniques, effectively revealing degradation mechanisms and trends, and providing a scientific basis for environmental protection and resource management. Attached Figure Description

[0053] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0054] Figure 1 This is a flowchart of a method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy, according to an embodiment of the present invention. Detailed Implementation

[0055] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0057] This embodiment proposes a method for identifying vegetation degradation in mining areas based on remote sensing spectra, such as... Figure 1 As shown, the specific steps include:

[0058] Collect multispectral image data of the mining area;

[0059] Based on multispectral image data of the mining area, principal component analysis was used to extract the main spectral feature vectors and determine the vegetation spectral feature set.

[0060] Based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels to obtain a vegetation degradation classification map.

[0061] The spatial distribution pattern is extracted from the vegetation degradation classification map, the area ratio and location coordinates of pixels in each category are calculated, and the spatial distribution characteristics of the degradation degree are determined.

[0062] Based on the spatial distribution characteristics of degradation, multi-temporal remote sensing image sequences are acquired, and changes in the classification map are tracked through temporal difference analysis to determine the dynamic trend.

[0063] By integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report is obtained to reveal the trend of vegetation degradation in mining areas.

[0064] Furthermore, the acquisition of multispectral image data of the mining area includes:

[0065] Preliminary multispectral image data of the mining area were collected using a satellite remote sensing platform;

[0066] Using preset band filtering rules, specific bands related to vegetation cover are extracted from the preliminary multispectral image data to obtain the filtered band dataset;

[0067] Based on the selected band dataset, an initial correction operation is performed to eliminate atmospheric interference and sensor bias, generating multispectral image data of the mining area.

[0068] Specifically, when collecting multispectral data of the mining area via satellite remote sensing platforms, high-resolution satellite sensors can be used to acquire image data covering the entire mining area, including information in multiple bands such as visible light and near-infrared. For processing vegetation-covered areas, assuming a mining area of ​​100 square kilometers and a data resolution of 10 meters, preprocessing can initially identify approximately 30% of the area as potentially covered by vegetation. This data acquisition method provides high-precision foundational information for subsequent analysis. In applying band selection rules, near-infrared and red bands can be prioritized for extraction, as these are most sensitive to the spectral reflectance characteristics of vegetation. The resulting dataset will primarily reflect the health status and cover density of the vegetation, reducing interference from other irrelevant bands and thus improving analysis efficiency. This selection helps focus on vegetation characteristics and avoids data redundancy. For initial correction operations, atmospheric correction models can be used to eliminate interference factors such as clouds and water vapor, while also correcting for sensor deviations caused by angle or hardware differences. Assuming the brightness deviation of the corrected image is reduced from the original 15% to 2%, the generated corrected base image more closely approximates the true surface reflectance. This correction significantly improves the accuracy of subsequent region classification. When dividing areas into vegetation-covered and non-vegetated regions, a threshold can be set based on the vegetation index value of the corrected image. Areas with an index higher than 0.3 are classified as vegetation-covered areas, and those lower are classified as non-vegetated areas. If the boundaries are blurred, the spectral differences between adjacent pixels can be analyzed. If the difference is less than 5%, they are classified into the same type, ultimately resulting in a clear region classification. This method effectively reduces misclassification and improves the accuracy of boundary division.

[0069] Furthermore, based on multispectral image data of the mining area, principal component analysis was used to extract the main spectral feature vectors, determining that the vegetation spectral feature set includes:

[0070] The multispectral image data of the mining area is standardized through data preprocessing to obtain a unified image data matrix;

[0071] Based on a unified image data matrix, principal component analysis is used to reduce the dimensionality of the spectral data, extract key spectral feature vectors, and determine the vegetation spectral feature set.

[0072] Specifically, when processing corrected multispectral image data, image standardization can be performed first to ensure that data from different sources have a consistent scale range. For example, if some bands in the acquired image data have grayscale values ​​ranging from 0 to 255, while others range from 0 to 1000, standardization can unify the numerical range of all bands to between 0 and 1. This facilitates subsequent analysis and avoids biases caused by scale differences. For the unified image data matrix, principal component analysis (PCA) can be used for dimensionality reduction, transforming high-dimensional spectral data into a few principal components. For instance, if the original data contains 10 bands, and analysis reveals that the first two principal components explain 90% of the variance, these two principal components can be extracted as key feature vectors, reducing data redundancy while retaining essential information. When separating reflectance components related to vegetation health, focus can be placed on reflectance data in specific bands. For example, if the difference in reflectance between the near-infrared and red bands is closely related to vegetation health, extracting data from these bands can construct indicators reflecting health status, laying the foundation for subsequent analysis.

[0073] Furthermore, based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, resulting in a vegetation degradation classification map including:

[0074] If the reflectance of the feature vector in the vegetation spectral feature set is lower than a preset threshold, obtain the judgment result;

[0075] Based on the judgment results, the potential interference area is obtained;

[0076] The mask filtering algorithm is used to remove noise caused by soil pollution and surface damage in potential interference areas, and a purified spectral feature set is obtained.

[0077] Based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, thus obtaining a vegetation degradation classification map.

[0078] Furthermore, by using a masking filtering algorithm to remove noise caused by soil pollution and surface damage in potential interference areas, the purified spectral feature set is obtained, including:

[0079] Based on the potential interference areas, obtain the interference area dataset;

[0080] A mask filtering algorithm was used to process the dataset of the interference area, and the noise components caused by soil pollution and surface damage were separated to obtain a preliminarily purified spectral dataset.

[0081] Based on the initially purified spectral dataset, the remaining abnormal reflectance points are detected. If abnormal points are detected, they are repaired by interpolation to obtain the repaired spectral dataset.

[0082] By smoothing the repaired spectral dataset to eliminate minor fluctuation noise, a purified spectral feature set is obtained.

[0083] Specifically, when identifying potential interference areas, if the reflectance value of a certain area is lower than a preset threshold, such as 0.2, it can be marked as a potential interference area. For example, if an area has a reflectance of only 0.15 in the 700 nm band, significantly lower than the reflectance level of normal vegetation, it may be due to exposed soil or sparse vegetation. In this case, its range needs to be marked and included in the interference area dataset for subsequent targeted processing. When using a mask filtering algorithm to process the interference area dataset, noise components caused by soil pollution or surface damage can be separated by setting specific spectral threshold ranges. For example, if a soil-polluted area has abnormally high reflectance in the near-infrared band, while a surface-damaged area exhibits low reflectance in the visible light band, mask filtering can effectively distinguish these noises, resulting in a preliminarily purified spectral dataset. This processing helps improve data purity and provides a more reliable foundation for subsequent analysis. When detecting and interpolating residual abnormal reflectance points, if the reflectance values ​​of some points suddenly deviate from the normal range—for example, if the reflectance of a point in a continuous band drops sharply from 0.3 to 0.1—interpolation can be performed using the average value of neighboring points to smooth out abnormal fluctuations. This method maintains the continuity of the spectral curve and avoids data distortion. When smoothing the repaired spectral dataset, a moving average can be used to eliminate minor fluctuation noise. Assuming the original data exhibits slight fluctuations in certain bands, smoothing by averaging five adjacent points can make the spectral curve smoother, reduce noise interference, and improve data readability. When the final purified spectral feature set is stored in a pre-defined database, a unique identifier can be generated for each region, such as using latitude and longitude coordinates combined with a timestamp, forming a coding format like "39.9N-116.3E-20231001". This method facilitates data management and traceability, ensuring that the available vegetation spectral dataset for a specific region can be quickly located during subsequent analysis. Through the coordinated processing of the above multiple stages, the quality and usability of spectral data can be significantly improved, providing strong support for the accurate assessment of vegetation status.

[0084] Furthermore, based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, resulting in a vegetation degradation classification map including:

[0085] The purified spectral feature set is standardized according to a preset format to obtain a structured spectral feature dataset.

[0086] A support vector machine classifier is trained using a structured spectral feature dataset to obtain a classification model;

[0087] By using a classification model, the health status of vegetation pixels is classified to obtain complete health status classification data;

[0088] Based on complete health status classification data, vegetation pixel degradation distribution information is generated. The classification data is mapped to geographical location information through spatial mapping technology, and a vegetation degradation classification map is output.

[0089] Specifically, when training a support vector machine (SVM) classifier for a structured spectral feature dataset, it can be based on labeled health status data. For example, 5000 pixels can be extracted from historical data, with 3000 labeled as healthy and 2000 labeled as varying degrees of degradation. By learning from these labeled data, the classifier can identify the feature boundaries between healthy and degraded states. During training, the model gradually optimizes its classification parameters based on the intensity and distribution of spectral feature values, forming a decision boundary capable of distinguishing different states. When using the trained classification model to classify the health status of vegetation pixels, assuming an analysis of 2000 pixels in a certain area, 1800 pixels have spectral feature values ​​within the normal range and are classified as normal. Of the remaining 200 pixels, 150 have feature values ​​slightly deviating from the normal range, meeting the criteria for mild degradation, and are therefore classified as mildly degraded. Further analysis of the remaining 50 pixels revealed that their feature values ​​significantly deviated from the normal range, meeting the criteria for severe degradation, and were ultimately classified as severely degraded. When generating vegetation pixel degradation distribution information, spatial mapping techniques can be used to map the classification data to geographical locations. Assuming a region has a 100x100 latitude and longitude grid, with each grid point corresponding to a pixel's classification result, mapping the classification data onto the grid provides a visual representation of the distribution of normal, mildly degraded, and severely degraded areas. This method helps to quickly locate severely degraded areas, providing a basis for subsequent management. For data storage and visualization, color coding can be used to distinguish different status areas. Assuming green represents normal status, yellow represents mild degradation, and red represents severe degradation, in the final distribution map, green areas account for 70%, yellow areas for 20%, and red areas for 10%. This visualization allows users to intuitively understand the vegetation health status, and storing this data in a database supports long-term monitoring. This approach significantly improves data utilization efficiency. The design of the entire process demonstrates that every step, from data processing and classification to visualization, is closely aligned with vegetation health assessment. Through multi-level classification and spatial mapping, a comprehensive understanding of vegetation distribution and degradation can be achieved, providing a reliable basis for relevant decision-making. This systematic approach not only improves data accuracy but also provides strong technical support for subsequent vegetation protection and management.

[0090] Furthermore, spatial distribution patterns are extracted from the vegetation degradation classification map, and the area ratio and location coordinates of pixels in each category are calculated to determine the spatial distribution characteristics of the degradation degree, including:

[0091] Based on the vegetation degradation classification map, the distribution information of pixels in each category is obtained by scanning pixel by pixel;

[0092] Based on the distribution information of pixels of each category, the pixel area and area ratio of each type of degradation are calculated, and the statistical results of each type of pixel are determined by the grid division method.

[0093] Based on the area proportion data in the statistical results, the position coordinates of pixels of each category are obtained, and the spatial mapping technology is used to determine the clustering of each category in the spatial distribution.

[0094] Based on the aggregation in the spatial distribution, the spatial pattern of vegetation degradation is analyzed. If the pixel aggregation degree of a certain type of degradation is higher than the preset threshold, it is marked as a high-density degradation area, and the spatial distribution characteristics are obtained.

[0095] Based on the location coordinates and distribution characteristics of high-density degraded areas, and combined with the original data of the classified images, the random forest algorithm is used to further verify the spatial pattern and determine the spatial distribution law of the degree of degradation.

[0096] Based on spatial distribution patterns, distribution characteristic data of degradation degree for each category are generated. Through data overlay processing, the spatial distribution characteristics of degradation degree are determined.

[0097] Specifically, when processing vegetation degradation pixel data in classified images, the image can be divided into multiple small regions by scanning pixel by pixel, and the pixels within each region can be labeled with a category. Assuming an image contains 1 million pixels, and scanning reveals that 30% of the pixels are labeled as mildly degraded, these pixels are mainly concentrated in the northwest corner of the image. This method helps to quickly locate the initial distribution of degradation, laying the foundation for subsequent analysis. When calculating the area and proportion of pixels in each category, a grid partitioning method can be used, dividing the entire image into 10x10 grid cells, each covering a certain area. Assuming each grid cell represents 1 square kilometer, statistical analysis shows that a certain type of degraded pixel accounts for 60% within a certain grid, indicating that there may be a relatively concentrated degradation phenomenon in that area. This method facilitates the quantitative analysis of the distribution scale of different degrees of degradation. For the spatial mapping of area proportion data, the spatial clustering of each category can be determined by extracting the pixel location coordinates. For example, if the coordinates of severely degraded pixels are concentrated within a 5-kilometer radius in a certain area, this high clustering may indicate significant local environmental pressure. Spatial mapping technology can intuitively reflect the distribution patterns of degradation categories, helping to identify key areas. When analyzing the spatial patterns of vegetation degradation, if the clustering degree of a certain type of degradation pixels exceeds a preset threshold, such as 80%, it can be marked as a high-density degradation area. For example, in a forest area, a large area of ​​pixels in the southeast is marked as high-density severely degraded, which may be related to over-logging or pests and diseases. This marking method helps to quickly pinpoint problem areas and improve the targeting of subsequent interventions. Further validation of high-density degradation areas can be achieved by combining the original classification image data and analyzing the spatial pattern using a random forest algorithm. For example, if during the validation process it is found that severely degraded pixels in a certain area are highly correlated with soil moisture data, this indicates that degradation may be related to water loss. This validation method can improve the reliability of distribution patterns and provide data support for subsequent research. When generating distribution characteristic data, data overlay processing can be used to integrate the spatial information of the degradation degree of each category. For example, if mild degradation is mainly distributed in the peripheral areas, while severe degradation is concentrated in the central area, this trend may reflect a gradient change in environmental pressure. Data overlay helps to reveal the overall distribution trend and improve the comprehensiveness of the analysis. When constructing a visual output of spatial distribution, heatmap technology can be used to represent different degrees of degradation with varying shades of color. For example, severely degraded areas are marked in dark red, and lightly degraded areas in light yellow. This intuitive presentation facilitates a quick understanding of the spatial characteristics of vegetation degradation. Generating heatmaps can effectively support decision-making and enhance the clarity of data representation.

[0098] Furthermore, by tracking changes in the classification map through temporal difference analysis, the dynamic trend of change can be determined, including:

[0099] The temporal difference method is used to perform layer-by-layer comparative analysis on multi-temporal remote sensing image sequences, calculate pixel-level differences between images, and determine the range of change areas.

[0100] Based on the range of the changed area, classify image data, divide the region according to the spatial distribution characteristics, and determine the evolution of the degree of degradation in each region.

[0101] Based on the evolution, obtain the dynamic trend of change.

[0102] Specifically, when using the temporal difference method for layer-by-layer comparative analysis, images from two adjacent years can be compared at the pixel level to identify areas where vegetation cover has decreased by more than 10% as the scope of change. The core of this method lies in capturing subtle changes over time, especially in areas with significant seasonal fluctuations, effectively distinguishing between natural fluctuations and long-term degradation trends. When acquiring categorized image data for the scope of change and combining it with spatial distribution characteristics for regional division, the change area can be divided into multiple small grids based on geographical units, with each grid area set to 1 square kilometer. By analyzing the distribution of degradation within each grid, for example, if the proportion of degraded pixels in a certain grid exceeds 50%, that area can be identified as a key degradation area. This division method helps to accurately locate problem areas. When extracting change trajectory data and analyzing the dynamic evolution direction, data from three consecutive years can be compared to observe whether the degree of degradation shows a continuously worsening trend. For example, if the vegetation cover of a certain area decreases by 5% in the first year, 8% in the second year, and 12% in the third year, it can be inferred that its evolution path is accelerated degradation. This trajectory analysis can provide data support for subsequent predictions. When using the Support Vector Machine (SVM) algorithm to classify and predict change trajectories, historical data can be divided into training and testing sets. The model is then trained to identify the probability distribution of future trends corresponding to different degradation paths. For example, predicting that a region has a 70% chance of worsening degradation within the next two years provides a reference for management decisions. This prediction method enhances the scientific rigor of trend judgment. When comprehensively assessing dynamic evolution trends, combining probability distribution and trend judgment logic, areas with a higher probability of worsening degradation (above 60%) can be marked as high-risk areas, while simultaneously analyzing the potential impact range of their surrounding areas. This assessment method helps to comprehensively understand the potential spread risk of degradation trends. When generating spatial distribution maps, high-risk areas can be marked with dark colors, and low-risk areas with light colors, creating an intuitive visual effect that aids in judging the potential change trajectory of degradation levels in different regions. This mapping method clearly presents spatial distribution characteristics, facilitating the development of targeted measures.

[0103] Furthermore, by integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report is obtained, including:

[0104] By mapping the monitoring data in the dynamic trend, a comprehensive assessment result of vegetation degradation in the mining area can be obtained;

[0105] Based on the comprehensive evaluation results and the analysis of the driving mechanisms, analytical data on the mechanism of vegetation degradation in mining areas are generated.

[0106] Based on the mechanism, analyze the data to obtain a comprehensive monitoring and identification report.

[0107] Specifically, by integrating long-term evolutionary trends with dynamic change data for multi-dimensional mapping, the degradation process can be visualized by region and time. For example, if the degradation rate in the eastern part of a mining area is relatively fast, with coverage decreasing from 60% to 40%, mapping can clearly show the problem area, providing an intuitive reference for management decisions. When generating data on the analysis of vegetation degradation mechanisms in mining areas, the overall trend can be judged by combining comprehensive assessment results. If the assessment shows that the degradation trend exceeds expectations, and the impact value of soil erosion as a key factor reaches 30%, its weight needs to be recalibrated, and the monitoring focus adjusted. This mechanism analysis helps optimize resource allocation and ensures that the analysis conclusions are more realistic. Through the above multi-faceted analysis and examples, it can be seen that each step from data classification to mechanism analysis is closely linked and progressively advanced, providing systematic support for monitoring vegetation degradation in mining areas. This method not only improves data utilization efficiency but also provides a scientific basis for the formulation of subsequent management measures, demonstrating high practical value.

[0108] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations 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. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying vegetation degradation in mining areas based on remote sensing spectra, characterized in that, include: Collect multispectral image data of the mining area; Based on the multispectral image data of the mining area, principal component analysis was used to extract the main spectral feature vectors and determine the vegetation spectral feature set. Based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels to obtain a vegetation degradation classification map. The spatial distribution pattern is extracted from the vegetation degradation classification map, the area ratio and location coordinates of pixels in each category are calculated, and the spatial distribution characteristics of the degradation degree are determined. Based on the spatial distribution characteristics of the degradation degree, multi-temporal remote sensing image sequences are acquired, and changes in the classification map are tracked through temporal difference analysis to determine the dynamic change trend; By integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report is obtained to reveal the trend of vegetation degradation in mining areas.

2. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, The multispectral image data collected from the mining area includes: Preliminary multispectral image data of the mining area were collected using a satellite remote sensing platform; Using preset band filtering rules, specific bands related to vegetation cover are extracted from the preliminary multispectral image data to obtain the filtered band dataset. Based on the filtered band dataset, an initial correction operation is performed to eliminate atmospheric interference and sensor bias, generating multispectral image data of the mining area.

3. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, Based on the multispectral image data of the mining area, principal component analysis was used to extract the main spectral feature vectors, and the vegetation spectral feature set was determined to include: The multispectral image data of the mining area is standardized through data preprocessing to obtain a unified image data matrix; Based on the unified image data matrix, principal component analysis is used to reduce the dimensionality of the spectral data, extract key spectral feature vectors, and determine the vegetation spectral feature set.

4. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, Based on the vegetation spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, resulting in a vegetation degradation classification map including: If the reflectance of the feature vectors in the vegetation spectral feature set is lower than a preset threshold, the judgment result is obtained; Based on the judgment results, the potential interference area is obtained; The noise caused by soil pollution and surface damage in the potential interference area is removed by a mask filtering algorithm to obtain a purified spectral feature set. Based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, thus obtaining the vegetation degradation classification map.

5. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 4, characterized in that, Noise caused by soil pollution and surface damage in the potential interference areas is removed using a masking filtering algorithm, resulting in a purified spectral feature set including: Based on the potential interference regions, obtain the interference region dataset; The interference region dataset was processed using a mask filtering algorithm to separate the noise components caused by soil pollution and surface damage, resulting in a preliminarily purified spectral dataset. Based on the preliminarily purified spectral dataset, the remaining abnormal reflectance points are detected. If abnormal points are detected, they are repaired by interpolation to obtain the repaired spectral dataset. By smoothing the repaired spectral dataset to eliminate minor fluctuation noise, the purified spectral feature set is obtained.

6. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 4, characterized in that, Based on the purified spectral feature set, a support vector machine classifier is used to classify the health status of vegetation pixels, distinguishing between normal, mildly degraded, and severely degraded categories, resulting in the vegetation degradation classification map, which includes: The purified spectral feature set is standardized according to a preset format to obtain a structured spectral feature dataset. The support vector machine classifier is trained using the structured spectral feature dataset to obtain a classification model; The health status of the vegetation pixels is classified using the classification model to obtain complete health status classification data. Based on the complete health status classification data, vegetation pixel degradation distribution information is generated. The classification data is mapped to geographical location information using spatial mapping technology, and the vegetation degradation classification map is output.

7. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, Extracting spatial distribution patterns from vegetation degradation classification maps, calculating the area ratio and location coordinates of pixels in each category, and determining the spatial distribution characteristics of degradation degree include: Based on the vegetation degradation classification map, the distribution information of pixels in each category is obtained by scanning pixels one by one; Based on the distribution information of pixels of each category, the pixel area and area ratio of each type of degradation are calculated, and the statistical results of each type of pixel are determined by the grid division method. Based on the area proportion data in the statistical results, the position coordinates of pixels of each category are obtained, and the clustering of each category in the spatial distribution is determined by spatial mapping technology. Based on the aggregation in the spatial distribution, the spatial pattern of vegetation degradation is analyzed. If the pixel aggregation degree of a certain type of degradation is higher than a preset threshold, it is marked as a high-density degradation area, and the spatial distribution characteristics are obtained. Based on the location coordinates and distribution characteristics of high-density degraded areas, and combined with the original data of the classified images, the random forest algorithm is used to further verify the spatial pattern and determine the spatial distribution law of the degree of degradation. Based on the spatial distribution pattern, distribution characteristic data of degradation degree for each category are generated. Through data overlay processing, the spatial distribution characteristics of degradation degree are determined.

8. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, By tracking changes in the classification map through time-series difference analysis, we can determine dynamic trends, including: The temporal difference method is used to perform a layer-by-layer comparative analysis on the multi-temporal remote sensing image sequence to calculate the pixel-level differences between images and determine the range of the change area. Based on the range of the changed area, classified image data is obtained, and the region is divided in combination with the spatial distribution characteristics to determine the evolution of the degree of degradation in each region. Based on the described evolution, the dynamic trend of change is obtained.

9. The method for identifying vegetation degradation in mining areas based on remote sensing spectroscopy according to claim 1, characterized in that, By integrating driving mechanism indicators from dynamic trends, a comprehensive monitoring and identification report is obtained, including: By mapping the monitoring data in the dynamic trend, a comprehensive assessment result of vegetation degradation in the mining area is obtained; Based on the comprehensive evaluation results and the analysis of the driving mechanism, analytical data on the mechanism of vegetation degradation in mining areas are generated. The data is analyzed according to the mechanism to obtain the comprehensive monitoring and identification report.

Images