Coal mining vegetation disturbance analysis method based on long-time-sequence multi-source data

Abstract

The invention discloses a coal mining vegetation disturbance analysis method based on long-time-sequence multi-source data, and relates to the technical field of mining area ecological monitoring and evaluation.The coal mining vegetation disturbance analysis method comprises the steps that on the basis of preprocessed mining area long-time-sequence spectral image data, a vegetation classification map is obtained through principal component analysis in combination with a classification algorithm; based on the vegetation classification map, obtaining a vegetation health index, evaluating the vegetation health condition and the vegetation change trend of the mining area, and obtaining vegetation health state time sequence data; based on the vegetation health state time sequence data, obtaining high-quality vegetation comprehensive data; constructing a vegetation state prediction model based on an LSTM model; and predicting a future vegetation state through a vegetation state prediction model based on the high-quality vegetation comprehensive data. According to the method, through an efficient multi-source data fusion mechanism and an advanced LSTM time sequence prediction technology, the accuracy and comprehensiveness of mining area vegetation state evaluation are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of mining area ecological monitoring and assessment, in particular to a coal mining vegetation disturbance analysis method based on long time series multi-source data. Background Art

[0002] In recent years, with the increasing frequency and scale of coal mining activities, the issue of ecological environment protection in mining areas has gradually attracted widespread attention. As an important part of the mining area ecosystem, the health of vegetation directly reflects the quality of the mining area environment and the effect of ecological restoration. Traditional vegetation monitoring methods mainly rely on ground surveys and static remote sensing image analysis. These methods have limitations in the timeliness and spatial coverage of data acquisition, and are difficult to meet the needs of modern mining area ecological management.

[0003] Although existing technologies have made some progress in vegetation monitoring in mining areas, there are still some shortcomings. First, existing technologies usually lack an effective fusion mechanism when processing long-term multi-source data. Various types of data come from different sources and have different temporal and spatial resolutions. How to effectively integrate these heterogeneous data to form a high-quality comprehensive data set is a major challenge facing current technology. For example, in practical applications, the temporal resolution of hyperspectral images is low, while the spatial resolution of meteorological data is limited. It is difficult to fully reflect the true state of vegetation in mining areas by relying solely on a single data source. Secondly, existing technologies often ignore the influence of the time dimension in the calculation of vegetation health index and status prediction. The health status of vegetation in mining areas is not only affected by current environmental factors, but also closely related to historical data. Most existing models fail to fully consider the long-term dependencies in time series data, resulting in the need to improve the accuracy and robustness of the prediction results. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a coal mining vegetation disturbance analysis method based on long time series multi-source data to solve the problems of insufficient long time series multi-source data fusion mechanism and neglect of the influence of time dimension in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, the present invention provides a coal mining vegetation disturbance analysis method based on long-time series multi-source data, which includes collecting long-time series spectral image data of the mining area and preprocessing the long-time series spectral image data of the mining area; based on the preprocessed long-time series spectral image data of the mining area, obtaining a vegetation classification map through principal component analysis and combining with a classification algorithm; based on the vegetation classification map, obtaining a vegetation health index, evaluating the health status of vegetation in the mining area and the vegetation change trend, and obtaining vegetation health status time series data; based on the vegetation health status time series data, obtaining high-quality vegetation comprehensive data; based on the LSTM model, constructing a vegetation status prediction model; based on the high-quality vegetation comprehensive data, predicting the future vegetation status through the vegetation status prediction model.

[0007] As a preferred solution of the coal mining vegetation disturbance analysis method based on long-time series multi-source data of the present invention, wherein: the long-time series spectral image data of the mining area includes multi-temporal images of the mining area, auxiliary multi-source data, and hyperspectral and hyperspectral images of the mining area; The preprocessing of the long-time spectral image data of the mining area includes radiation calibration, atmospheric correction and geometric correction.

[0008] As a preferred solution of the coal mining vegetation disturbance analysis method based on long-time series multi-source data described in the present invention, wherein: based on the pre-processed long-time series spectral image data of the mining area, the vegetation classification map is obtained by principal component analysis combined with a classification algorithm. The specific process is as follows: Convert the preprocessed long-time series spectral image data of the mining area into a data matrix; Normalize each band in the data matrix; Based on the standardized data matrix, obtain the covariance matrix; Calculate eigenvalues ​​and eigenvectors based on the covariance matrix to obtain a new base layer; Project the preprocessed long-time spectral image data of the mining area onto the new base layer to obtain dimension-reduced data; Based on the dimension reduction data, the spectral characteristics of vegetation are obtained; Based on the vegetation spectral characteristics, the vegetation spectral characteristic intensity of each data sample point is calculated, and the expression is: ; in, It is The vegetation spectral characteristic intensity of the dimension-reduced data of the data sample points, It is The dimension reduction data of data sample points, It is The weight coefficients of the principal components, It is the control The control parameter for the decay speed of the first exponential function on the principal component is It is The center position of the first exponential function on the principal component, It is the control The control parameter for the decay speed of the second exponential function on the principal component is It is The central position of the second exponential function on the principal component, It is The dimension reduction data of the data sample points is in The values ​​on the principal components, is the total number of principal components, is the index variable of the principal component, is the index variable of the data sample point; Define the vegetation distinction threshold based on the vegetation spectral characteristic intensity distribution ; Analyze vegetation areas based on vegetation spectral characteristic intensity and vegetation distinction threshold; when When , it is marked as a vegetation area; when When , it is marked as non-vegetated area; Based on the classification results, morphological operations and connected domain analysis are used for optimization, and adjacent areas of the same type are merged to obtain the vegetation classification map.

[0009] As a preferred solution of the coal mining vegetation disturbance analysis method based on long time series multi-source data described in the present invention, wherein: the vegetation health index is obtained based on the vegetation classification map, and the specific steps are as follows: Based on the vegetation classification map, vegetation sample points are obtained through spatial clustering algorithm; The vegetation health index is calculated through vegetation sample points and vegetation spectral characteristic intensity. The expression is: ; in, It is Vegetation sample points at time The vegetation health index, It is Vegetation sample points, It's in time Dependence The weight coefficients of the principal components, It is On the principal component The weight coefficient of It is The comprehensive spectral index of vegetation sample points, It is On the principal component The weight coefficient of It is the control The parameters of the time decay rate of the principal components, It is The temporal center position of the principal component, is the control factor in the denominator The parameter of the time decay rate of the principal component, is the denominator The temporal center position of the principal component, is the index variable of the principal component, is the index variable of the vegetation sample point.

[0010] As a preferred solution of the coal mining vegetation disturbance analysis method based on long time series multi-source data described in the present invention, wherein: the health status of vegetation in the mining area and the vegetation change trend are evaluated, and the vegetation health status time series data is obtained. The specific steps are as follows: By analyzing the distribution of historical vegetation health indices, health thresholds are defined and damage threshold ; Based on vegetation health index, health threshold and damage threshold , determine the health status of vegetation; when When , it means the vegetation is in a healthy state; when When , it means that the vegetation is in sub-health condition; when When , it means that the vegetation is in a damaged state; The classification results and the corresponding vegetation health index at each time point are sorted in chronological order to obtain time series data; The vegetation health index in the time series data is analyzed by time series decomposition method to obtain the vegetation change trend; Combine vegetation change trends with vegetation classification maps to obtain vegetation health status time series data.

[0011] As a preferred solution of the coal mining vegetation disturbance analysis method based on long time series multi-source data described in the present invention, wherein: based on the vegetation health status time series data, high-quality vegetation comprehensive data is obtained, and the specific steps are as follows: Temporally and spatially match auxiliary multi-source data with vegetation health status time series data; The vegetation health status time series data is fused with auxiliary multi-source data through Bayesian fusion method to obtain a multi-source fusion data set; Based on multi-source fusion data sets, high-quality vegetation comprehensive data are obtained through dynamic spatiotemporal correction algorithm.

[0012] As a preferred solution of the coal mining vegetation disturbance analysis method based on long time series multi-source data described in the present invention, wherein: based on the LSTM model, a vegetation state prediction model is constructed, and the specific process is as follows: The historical high-quality vegetation comprehensive data at each time point are integrated by weighted average method to extract the historical vegetation feature vector; Based on the historical vegetation feature vector, the time series samples are constructed by the sliding window method; The LSTM model is used as the basic model; The input layer receives time series samples; The LSTM layer captures long-term dependencies by stacking multiple LSTM layers; Add a Dropout layer to the LSTM layer to randomly discard some neurons; The fully connected layer extracts features and reduces the dimension based on the output of the LSTM layer and maps it to the output layer; Output the prediction results through the output layer; Finally, a vegetation status prediction model was constructed.

[0013] As a preferred solution of the coal mining vegetation disturbance analysis method based on long-term multi-source data described in the present invention, the future vegetation state is predicted based on high-quality vegetation comprehensive data through a vegetation state prediction model. The specific steps are as follows: The vegetation feature vectors in high-quality vegetation comprehensive data are extracted by weighted average method; Based on the vegetation feature vector, the predicted vegetation state value at the future prediction target time point is calculated, and the expression is: ; in, is the predicted vegetation state value at the future prediction target time point, It is The weight coefficient of the LSTM model output in the time window, is the mapping function of the LSTM model, is the future prediction target time point The vegetation feature vector, It is The weight coefficient of the external influencing factor output of the samples in the time window, is the future prediction target time point The characteristic vector of the external influencing factors, It is the control The time decay rate parameter of samples in a time window, is the future prediction target time point, It is The reference time point of the samples in the time window, is the control factor in the denominator The parameter of the speed at which the sample weight decays over time in a time window, is the number of samples in the time window, is the index variable of the samples in the time window; Define vegetation status thresholds based on historical vegetation health data ; Analyze the future vegetation state based on the predicted vegetation state value and vegetation state threshold at the future prediction target time point; when When , it means that the future vegetation status is expected to be good; when When the number is 250, it means that the vegetation status is facing degradation in the future, which requires attention and measures to be taken.

[0014] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of a coal mining vegetation disturbance analysis method based on long time series multi-source data as described in the first aspect of the present invention.

[0015] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of a method for analyzing vegetation disturbance in coal mining based on long time series multi-source data as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: the present invention significantly improves the accuracy and comprehensiveness of the assessment of vegetation status in mining areas through an efficient multi-source data fusion mechanism and advanced LSTM time series prediction technology, while enhancing the foresight and scientific nature of the prediction of future vegetation status, thereby providing reliable data support and effective preventive measures for ecological management and environmental protection in mining areas, and ensuring the sustainable development of mining areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a flow chart of a method for analyzing vegetation disturbance in coal mining based on long-term multi-source data in Example 1.

[0019] Figure 2 This is a flow chart for obtaining high-quality vegetation comprehensive data in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data, including the following steps: S1. The long-term spectral image data of the mining area includes multi-temporal images of the mining area, auxiliary multi-source data, and hyperspectral and hyperspectral images of the mining area.

[0024] S1.1. Preprocessing of long-term spectral image data of mining areas includes radiation calibration, atmospheric correction and geometric correction.

[0025] Specifically, radiometric calibration ensures the quantitative accuracy of data by converting the digital values ​​of images into physical radiance or reflectance; Atmospheric correction restores the true reflectance spectral characteristics of the surface by eliminating atmospheric scattering and absorption effects; Geometric correction corrects the geometric distortion in images so that images at different time points are accurately aligned in space, ensuring the accuracy of geographic positioning.

[0026] S2. Based on the preprocessed long-time spectral image data of the mining area, the vegetation classification map is obtained through principal component analysis combined with classification algorithm.

[0027] S2.1. Convert the preprocessed long-time series spectral image data of the mining area into a data matrix.

[0028] Furthermore, the preprocessed long-time spectral image data of the mining area are reorganized according to the time series and band order and converted into a three-dimensional or multi-dimensional data matrix.

[0029] S2.2. Standardize each band in the data matrix.

[0030] Furthermore, the mean and standard deviation of each band are calculated, and the pixel values ​​of each band are converted into standardized values ​​with zero mean and unit variance.

[0031] S2.3. Based on the standardized data matrix, obtain the covariance matrix.

[0032] Furthermore, the covariance between each band is calculated to form a square matrix reflecting the linear relationship between the bands, in which each element represents the covariance value between two bands.

[0033] S2.4. Calculate eigenvalues ​​and eigenvectors based on the covariance matrix to obtain a new base layer.

[0034] Furthermore, all eigenvalues ​​of the covariance matrix and their corresponding eigenvectors are calculated through mathematical methods (such as power iteration method, QR decomposition or Jacobi method) (eigenvalues ​​represent the variance of the data in the direction of the corresponding eigenvector, and eigenvectors define the new principal component direction), and a set of eigenvalues ​​and their corresponding eigenvectors are obtained. These are sorted by eigenvalue size to form a main eigenvector matrix, and the main eigenvectors are selected to form a new basis.

[0035] S2.5. Project the preprocessed long-time spectral image data of the mining area onto the new base layer to obtain dimension-reduced data.

[0036] Furthermore, the main eigenvector matrix is ​​used to perform a linear transformation on the original data and project it onto a new basis to obtain the reduced-dimensional data.

[0037] S2.6. Obtain vegetation spectral characteristics based on the dimensionality reduction data.

[0038] Furthermore, by analyzing the reduced-dimensional data, characteristic bands related to vegetation are extracted or vegetation indices are constructed to obtain spectral features that can characterize the health status and distribution of vegetation.

[0039] S2.7. Based on the vegetation spectral characteristics, the vegetation spectral characteristic intensity of each data sample point is calculated. The expression is: ; in, It is The vegetation spectral characteristic intensity of the dimension-reduced data of the data sample points, It is The dimension reduction data of data sample points, It is The weight coefficients of the principal components, It is the control The control parameter for the decay speed of the first exponential function on the principal component is It is The center position of the first exponential function on the principal component, It is the control The control parameter for the decay speed of the second exponential function on the principal component is It is The central position of the second exponential function on the principal component, It is The dimension reduction data of the data sample points is in The values ​​on the principal components, is the total number of principal components, is the index variable of the principal component, is the index variable of the data sample point.

[0040] It should be noted that the data sample point refers to each data point with specific geographical location and spectral information extracted from the preprocessed spectral image data.

[0041] S2.8. Define the vegetation distinction threshold based on the vegetation spectral characteristic intensity distribution .

[0042] S2.9, analyze vegetation area based on vegetation spectral characteristic intensity and vegetation distinction threshold; when When , it is marked as a vegetation area; when , it is marked as a non-vegetated area.

[0043] It should be noted that the spectral characteristic intensity values ​​of vegetation for all sample points are calculated and their histograms or cumulative distribution functions (CDFs) are plotted to visually display the data distribution; secondly, based on the distribution characteristics (such as the valley value in the bimodal distribution, a specific percentile, or through an automatic threshold algorithm such as the Otsu method), a threshold that can effectively distinguish between vegetation and non-vegetation areas is determined.

[0044] S2.10. Based on the classification results, morphological operations and connected domain analysis are used for optimization, and adjacent areas of the same type are merged to obtain a vegetation classification map.

[0045] S3. Obtain vegetation health index based on vegetation classification map.

[0046] S3.1. Based on the vegetation classification map, vegetation sample points are obtained through spatial clustering algorithm.

[0047] More specifically, through spatial clustering algorithms (such as K-means or DBSCAN), pixels with similar spectral characteristics are automatically identified and aggregated to form vegetation clusters, from which vegetation sample points are extracted.

[0048] S3.2. Calculate the vegetation health index through vegetation sample points and vegetation spectral characteristic intensity. The expression is: ; in, It is Vegetation sample points at time The vegetation health index, It is Vegetation sample points, It's in time Dependence The weight coefficients of the principal components, It is On the principal component The weight coefficient of It is The comprehensive spectral index of vegetation sample points, It is On the principal component The weight coefficient of It is the control The parameters of the time decay rate of the principal components, It is The temporal center position of the principal component, is the control factor in the denominator The parameter of the time decay rate of the principal component, is the denominator The temporal center position of the principal component, is the index variable of the principal component, is the index variable of the vegetation sample point.

[0049] It should be noted that vegetation sample points refer to the spatial locations that have been confirmed as vegetation categories and their corresponding spectral characteristics; It should also be noted that by calculating the weighted average of the spectral values ​​of each vegetation sample point in multiple bands or other comprehensive indicators (such as the normalized difference vegetation index NDVI), a comprehensive spectral index that can characterize the overall spectral characteristics of the sample point is formed.

[0050] S4. Evaluate the vegetation health status and vegetation change trends in the mining area and obtain vegetation health status time series data.

[0051] S4.1. Define health thresholds by analyzing the distribution of historical vegetation health indices and damage threshold .

[0052] It should be noted that by calculating the statistics of the historical vegetation health index (such as mean, standard deviation) and combining it with a histogram or cumulative distribution function, identifying natural groupings in the data or using an automatic threshold algorithm (such as the Otsu method), the critical value that can accurately divide different health levels can be set.

[0053] S4.2. Based on vegetation health index and health threshold and damage threshold , determine the health status of vegetation; when When , it means the vegetation is in a healthy state; when When , it means that the vegetation is in sub-health condition; when When , it indicates that the vegetation is in a damaged condition.

[0054] S4.3. Sort the classification results and the corresponding vegetation health index at each time point in chronological order to obtain time series data.

[0055] S4.4. Analyze the vegetation health index in the time series data through the time series decomposition method to obtain the vegetation change trend.

[0056] It should be noted that the time series decomposition technology (such as STL decomposition) is used to decompose the time series data of the vegetation health index into trend components, seasonal components and residual components, so as to intuitively extract and analyze the long-term change trend of vegetation health status.

[0057] S4.5. Combine vegetation change trends and vegetation classification maps to obtain vegetation health status time series data.

[0058] It should be noted that by superimposing the vegetation change trend and the vegetation classification map, the health status of different vegetation areas at each time point is analyzed to generate vegetation health status time series data containing time and space information.

[0059] S5. Obtain high-quality vegetation comprehensive data based on vegetation health status time series data.

[0060] S5.1. Temporally and spatially match the auxiliary multi-source data with the vegetation health status time series data.

[0061] It should be noted that the auxiliary multi-source data includes meteorological data, soil data and terrain data.

[0062] S5.2. The vegetation health status time series data is fused with the auxiliary multi-source data through the Bayesian fusion method to obtain a multi-source fusion data set.

[0063] It should be noted that the Bayesian fusion method is used to integrate the time series data of vegetation health status and auxiliary information from different sources, express uncertainty in the form of probability, and update the posterior distribution based on prior knowledge and observational data, so as to construct a comprehensive multi-source fusion dataset.

[0064] S5.3. Based on multi-source fusion data sets, high-quality vegetation comprehensive data are obtained through dynamic spatiotemporal correction algorithm.

[0065] It should be noted that the dynamic spatiotemporal correction algorithm is used to adjust the spatiotemporal inconsistencies in the fused dataset, eliminate biases and optimize data quality, thereby generating high-quality vegetation comprehensive data that accurately reflects the status of vegetation.

[0066] S6. Based on the LSTM model, a vegetation status prediction model is constructed.

[0067] S6.1. Integrate the historical high-quality vegetation comprehensive data at each time point through the weighted average method to extract the historical vegetation feature vector.

[0068] It should be noted that different weights are assigned according to the importance or relevance of the historical high-quality vegetation comprehensive data at each time point, and the multi-source data (such as spectral images, meteorological data, etc.) at each time point are weighted, and their weighted average values ​​are calculated. These weighted average values ​​are combined into a comprehensive data vector, which can fully reflect the vegetation status at that time point, extract key features from the comprehensive data vector, and form a historical vegetation feature vector; It should also be noted that key features refer to those attributes that can best characterize the state and health of vegetation and are crucial for predicting future vegetation changes, such as spectral indices (NDVI, EVI), meteorological conditions (temperature, precipitation), soil properties (moisture, nutrients), and topographic factors (elevation, slope); The method for determining key features is to identify the attributes that have the greatest impact on vegetation health by combining machine learning feature selection algorithms (such as recursive feature elimination, feature importance scoring), statistical analysis (such as correlation analysis, principal component analysis) and domain expertise, and ensure the effectiveness of these features in the prediction model through cross-validation.

[0069] S6.2. Based on the historical vegetation feature vector, the time series samples are constructed by the sliding window method.

[0070] It should be noted that a sliding window of fixed length is used to select feature vectors of consecutive time points in sequence to form a multidimensional sample matrix containing multiple time steps, thereby constructing a time series sample.

[0071] S6.3, based on LSTM model; The input layer receives time series samples; The LSTM layer captures long-term dependencies by stacking multiple LSTM layers; Add a Dropout layer to the LSTM layer to randomly discard some neurons; The fully connected layer extracts features and reduces the dimension based on the output of the LSTM layer and maps it to the output layer; Output the prediction results through the output layer; Finally, a vegetation status prediction model was constructed.

[0072] It should be noted that the benefits of building a vegetation state prediction model based on the LSTM model are: LSTM (Long Short-Term Memory Network) can effectively capture dependencies and complex patterns in long time series data, and is particularly suitable for processing vegetation state data with time dependence. By learning the temporal evolution of historical vegetation feature vectors, the LSTM model can more accurately predict future vegetation states and provide forward-looking ecological management recommendations, and its powerful nonlinear modeling capabilities help improve the accuracy and stability of predictions, thereby providing a scientific basis for ecological restoration and environmental protection in mining areas.

[0073] S7. Based on high-quality vegetation comprehensive data, the future vegetation status is predicted through the vegetation status prediction model.

[0074] S7.1. Extract vegetation feature vectors from high-quality vegetation comprehensive data using the weighted average method.

[0075] S7.2. Based on the vegetation feature vector, the predicted vegetation state value at the future prediction target time point is calculated. The expression is: ; in, is the predicted vegetation state value at the future prediction target time point, It is The weight coefficient of the LSTM model output in the time window, is the mapping function of the LSTM model, is the future prediction target time point The vegetation feature vector, It is The weight coefficient of the external influencing factor output of the samples in the time window, is the future prediction target time point The characteristic vector of the external influencing factors, It is the control The time decay rate parameter of samples in a time window, is the future prediction target time point, It is The reference time point of the samples in the time window, is the control factor in the denominator The parameter of the speed at which the sample weight decays over time in a time window, is the number of samples in the time window, is the index variable of the samples within the time window.

[0076] It should be noted that the vegetation feature vector at the future prediction target time point , through the existing long-span spectral image data and other auxiliary multi-source data, data analysis techniques (such as principal component analysis, machine learning models, etc.) are used to identify the features closely related to vegetation health, and the time series model constructed using these features is used for extrapolation or prediction; It should also be noted that external influencing factors are collected through a variety of channels: meteorological data such as temperature and precipitation are provided by weather stations and satellite remote sensing; soil monitoring networks measure soil moisture, temperature and chemical composition; environmental monitoring agencies record air quality and pollutant levels; The feature vector of external influencing factors at the future prediction target time point , by integrating data from multiple sources, specifically through standardization and time alignment, to ensure that the external influencing factors at each time point can be accurately represented. Finally, these processed data are combined into feature vectors.

[0077] S7.3. Define vegetation status thresholds based on historical vegetation health data .

[0078] S7.4, analyzing the future vegetation state based on the predicted vegetation state value and the vegetation state threshold at the future prediction target time point; when When , it means that the future vegetation status is expected to be good; when When the number is 250, it means that the vegetation status is facing degradation in the future, which requires attention and measures to be taken.

[0079] It should be noted that the vegetation status threshold is obtained by identifying natural grouping points in the data through statistical analysis (such as histograms, cumulative distribution functions or automatic threshold algorithms) and selecting critical values ​​that maximize the differences between categories and minimize the differences within categories.

[0080] It should also be noted that measures taken can include strengthening monitoring, implementing ecological restoration projects, and adjusting management strategies to promote vegetation recovery and protection.

[0081] This embodiment also provides a computer device, which is suitable for a coal mining vegetation disturbance analysis method based on long time series multi-source data, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement a coal mining vegetation disturbance analysis method based on long time series multi-source data as proposed in the above embodiment.

[0082] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0083] The present embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for analyzing vegetation disturbance in coal mining based on long-time series multi-source data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage device, a flash memory, a disk or an optical disk.

[0084] In summary, the present invention significantly improves the accuracy and comprehensiveness of the vegetation status assessment in mining areas through: efficient multi-source data fusion mechanism and advanced LSTM time series prediction technology, while enhancing the foresight and scientific nature of the prediction of future vegetation status, thereby providing reliable data support and effective preventive measures for ecological management and environmental protection in mining areas, and ensuring the sustainable development of mining areas.

[0085] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a coal mining vegetation disturbance analysis method based on long time series multi-source data is provided.

[0086] In order to verify the effectiveness and superiority of the above-mentioned mining area vegetation health status monitoring and prediction system (hereinafter referred to as "this system"), a typical mining area was selected as the research object. The mining area is located in northern China and has typical temperate continental climate characteristics, with an annual precipitation of about 300 mm and significant seasonal changes. Mining activities in the mining area have been going on for many years, which has caused varying degrees of impact on the local ecological environment, especially the changes in vegetation coverage and health status.

[0087] First, in order to verify the effectiveness and superiority of the mining area vegetation health status monitoring and prediction system, a northern mining area with typical temperate continental climate characteristics was selected as the research object. Long-term spectral image data from 2015 to 2023 were collected and preprocessed, including multi-temporal images, hyperspectral and hyperspectral images, and auxiliary multi-source data. These data were calibrated by radiation, atmospheric correction, and geometric correction to ensure the quality and accuracy of the data, providing a reliable basis for subsequent analysis.

[0088] Secondly, based on the preprocessed spectral image data, a series of mathematical transformations such as principal component analysis, standardization, covariance matrix calculation, eigenvalue and eigenvector calculation were used to obtain the reduced dimension data and extract the vegetation spectral characteristics from it. The vegetation distinction threshold was defined according to the distribution of vegetation spectral characteristic intensity, and the automatic identification of vegetation areas was realized. At the same time, based on the vegetation classification map, combined with the spatial clustering algorithm and the spectral characteristic intensity of the vegetation sample points, the vegetation health index was calculated to more comprehensively reflect the health status of the vegetation, and healthy and damaged thresholds were set to judge the vegetation status.

[0089] Next, in order to evaluate the difference between this system and existing technologies, the traditional ground survey method and the vegetation index analysis method based on a single remote sensing image were selected as the control group, and independent operations were performed on different blocks in the same mining area to ensure the consistency and fairness of the comparison. During the experiment, each method was subjected to detailed performance tests, including comparisons of vegetation coverage, health index, prediction accuracy, and time efficiency, and the corresponding data results were recorded and analyzed.

[0090] Finally, the vegetation state prediction model was constructed using the LSTM model. By training the time series samples of high-quality vegetation comprehensive data, long-term dependencies were captured to achieve accurate prediction of future vegetation states. Considering that external factors may affect the vegetation state, the system also integrates an auxiliary model to deal with these factors, making the prediction results more accurate and reliable. Through the above process, it is proved that this system has significant advantages in vegetation monitoring and prediction, and provides strong technical support for ecological restoration in mining areas.

[0091] The details are shown in Table 1 below: Table 1 Comparison of mining area vegetation analysis methods

[0092] Through the data analysis of the above table, it can be clearly seen that the mining area vegetation health status monitoring and prediction system has significant advantages over the traditional ground survey method and the vegetation index analysis method based on a single remote sensing image. The present invention integrates advanced data processing technology and machine learning models, such as LSTM models to capture the integration of long-term dependencies and external influencing factors, to achieve higher vegetation coverage recognition (82.3% vs 69.7% and 65.4%), more accurate vegetation health assessment (health index 0.89 vs 0.76 and 0.71), and significantly improved prediction accuracy (92.5% vs 78.3% and 74.1%). In addition, the efficiency of the system is also reflected, and the monitoring task can be completed in only 12 hours / year, while the traditional methods require 235 hours and 115 hours respectively. These advantages enable the present invention to provide more scientific, fast and accurate technical support in mining area ecological restoration and environmental protection.

[0093] This invention significantly improves the accuracy and efficiency of vegetation health status monitoring in mining areas by integrating advanced data processing technology and machine learning models, achieving a wider range of vegetation coverage identification, more accurate health status assessment, and high-precision prediction of future vegetation status. Compared with traditional methods, this invention greatly reduces the time required for monitoring and can effectively integrate a variety of external influencing factors, providing scientific, fast and reliable technical support for ecological restoration in mining areas, thereby promoting the goals of environmental protection and sustainable development.

[0094] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A coal mining vegetation disturbance analysis method based on long-term multi-source data, characterized by: include, Collect long-term spectral image data of the mining area and pre-process the long-term spectral image data of the mining area; Based on the pre-processed long-time spectral image data of the mining area, the vegetation classification map is obtained through principal component analysis combined with classification algorithm; Based on the vegetation classification map, obtain the vegetation health index, evaluate the vegetation health status and vegetation change trend in the mining area, and obtain the vegetation health status time series data; Obtain high-quality vegetation comprehensive data based on vegetation health status time series data; Based on the LSTM model, a vegetation status prediction model is constructed; Based on high-quality comprehensive vegetation data, the future vegetation status is predicted through the vegetation status prediction model.

2. The method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data according to claim 1, characterized in that: The long-term spectral image data of the mining area includes multi-temporal images of the mining area, auxiliary multi-source data, and high-spectral and hyper-spectral images of the mining area; The preprocessing of the long-time spectral image data of the mining area includes radiation calibration, atmospheric correction and geometric correction.

3. The method for analyzing vegetation disturbance in coal mining based on long-term multi-source data according to claim 2, characterized in that: Based on the pre-processed long-time spectral image data of the mining area, the vegetation classification map is obtained by principal component analysis combined with a classification algorithm. The specific process is as follows: Convert the preprocessed long-time series spectral image data of the mining area into a data matrix; Normalize each band in the data matrix; Based on the standardized data matrix, obtain the covariance matrix; Calculate eigenvalues ​​and eigenvectors based on the covariance matrix to obtain a new base layer; Project the preprocessed long-time spectral image data of the mining area onto the new base layer to obtain dimension-reduced data; Obtain vegetation spectral characteristics based on dimension reduction data; Based on the vegetation spectral characteristics, the vegetation spectral characteristic intensity of each data sample point is calculated, and the expression is: ； in, It is The vegetation spectral characteristic intensity of the dimension-reduced data of the data sample points, It is The dimension reduction data of data sample points, It is The weight coefficients of the principal components, It is the control The control parameter for the decay speed of the first exponential function on the principal component is It is The center position of the first exponential function on the principal component, It is the control The control parameter for the decay speed of the second exponential function on the principal component is It is The central position of the second exponential function on the principal component, It is The dimension reduction data of the data sample points is in The values ​​on the principal components, is the total number of principal components, is the index variable of the principal component, is the index variable of the data sample point; Define the vegetation distinction threshold based on the vegetation spectral characteristic intensity distribution ; Analyze vegetation areas based on vegetation spectral characteristic intensity and vegetation distinction threshold; when When , it is marked as a vegetation area; when When , it is marked as non-vegetated area; Based on the classification results, morphological operations and connected domain analysis are used for optimization, and adjacent areas of the same type are merged to obtain the vegetation classification map.

4. The method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data according to claim 3, characterized in that: The specific steps of obtaining the vegetation health index based on the vegetation classification map are as follows: Based on the vegetation classification map, vegetation sample points are obtained through spatial clustering algorithm; The vegetation health index is calculated through vegetation sample points and vegetation spectral characteristic intensity. The expression is: ； in, It is Vegetation sample points at time The vegetation health index, It is Vegetation sample points, It's in time Dependence The weight coefficients of the principal components, It is On the principal component The weight coefficient of It is The comprehensive spectral index of vegetation sample points, It is On the principal component The weight coefficient of It is the control The parameters of the time decay rate of the principal components, It is The temporal center position of the principal component, is the control factor in the denominator The parameter of the time decay rate of the principal component, is the denominator The temporal center position of the principal component, is the index variable of the principal component, is the index variable of the vegetation sample point.

5. The method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data according to claim 4, characterized in that: The specific steps of evaluating the vegetation health status and vegetation change trend in the mining area and obtaining vegetation health status time series data are as follows: By analyzing the distribution of historical vegetation health indices, health thresholds are defined and damage threshold ; Based on vegetation health index, health threshold and damage threshold , determine the health status of vegetation; when When , it means the vegetation is in a healthy state; when When , it means that the vegetation is in sub-health condition; when When , it means that the vegetation is in a damaged state; The classification results and the corresponding vegetation health index at each time point are sorted in chronological order to obtain time series data; The vegetation health index in the time series data is analyzed by time series decomposition method to obtain the vegetation change trend; Combine vegetation change trends with vegetation classification maps to obtain vegetation health status time series data.

6. The method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data according to claim 5, characterized in that: The specific steps of obtaining high-quality vegetation comprehensive data based on vegetation health status time series data are as follows: Temporally and spatially match auxiliary multi-source data with vegetation health status time series data; The vegetation health status time series data is fused with auxiliary multi-source data through Bayesian fusion method to obtain a multi-source fusion data set; Based on multi-source fusion data sets, high-quality vegetation comprehensive data are obtained through dynamic spatiotemporal correction algorithm.

7. The method for analyzing vegetation disturbance caused by coal mining based on long-term multi-source data according to claim 6, characterized in that: The vegetation state prediction model is constructed based on the LSTM model. The specific process is as follows: The historical high-quality vegetation comprehensive data at each time point are integrated by weighted average method to extract the historical vegetation feature vector; Based on the historical vegetation feature vector, the time series samples are constructed by the sliding window method; The LSTM model is used as the basic model; The input layer receives time series samples; The LSTM layer captures long-term dependencies by stacking multiple LSTM layers; Add a Dropout layer to the LSTM layer to randomly discard some neurons; The fully connected layer extracts features and reduces the dimension based on the output of the LSTM layer and maps it to the output layer; Output the prediction results through the output layer; Finally, a vegetation status prediction model was constructed.

8. The method for analyzing vegetation disturbance in coal mining based on long-term multi-source data according to claim 7, characterized in that: The method of predicting the future vegetation state based on high-quality vegetation comprehensive data through a vegetation state prediction model is as follows: The vegetation feature vectors in high-quality vegetation comprehensive data are extracted by weighted average method; Based on the vegetation feature vector, the predicted vegetation state value at the future prediction target time point is calculated, and the expression is: ； in, is the predicted vegetation state value at the future prediction target time point, It is The weight coefficient of the LSTM model output in the time window, is the mapping function of the LSTM model, is the future prediction target time point The vegetation feature vector, It is The weight coefficient of the external influencing factor output of the samples in the time window, is the future prediction target time point The characteristic vector of the external influencing factors, It is the control The time decay rate parameter of samples in a time window, is the future prediction target time point, It is The reference time point of the samples in the time window, is the control factor in the denominator The parameter of the speed at which the sample weight decays over time in a time window, is the number of samples in the time window, is the index variable of the samples in the time window; Define vegetation status thresholds based on historical vegetation health data ; Analyze the future vegetation state based on the predicted vegetation state value and vegetation state threshold at the future prediction target time point; when When , it means that the future vegetation status is expected to be good; when When the number is 250, it means that the vegetation status is facing degradation in the future, which requires attention and measures to be taken.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a method for analyzing vegetation disturbance in coal mining based on long-time series multi-source data as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for analyzing vegetation disturbance in coal mining based on long-time series multi-source data as described in any one of claims 1 to 8 are implemented.

Images