Ecological Environment Quality Prediction Method and System Based on Temporal Remote Sensing Images and Residual Reconstruction

Abstract

The present invention discloses an ecological environment quality prediction method and model based on time-series remote sensing images and residual reconstruction. The method includes: S1, obtaining an ecological environment quality sample data set, and dividing the ecological environment quality sample data set into a data subset only affected by climate factors and a data subset affected by both climate factors and human activities; S2, a multiple regression model adds the trained regression function A and regression function B in a raster map to form a comprehensive ecological environment quality prediction function model; S3, collecting climate factor data and human activity data in the research mining area, inputting them into the multiple regression model, and outputting the ecological environment quality prediction result. The present invention uses the ecological environment quality sample data set for residual reconstruction, trains the regression function A and regression function B respectively, constructs a comprehensive ecological environment quality prediction function model that comprehensively considers natural factors and human activities, and effectively realizes the prediction of the future ecological environment quality of the mining area.

Description

Technical Field

[0001] The present invention relates to the field of prediction of the ecological environment quality in mining areas, and particularly to a method and system for predicting the ecological environment quality based on time-series remote sensing images and residual reconstruction. Background Art

[0002] Mineral resources are important energy sources in our country and are of great significance for promoting the economic development of our country. However, during the process of mine exploitation, the land resources in the mining area will be occupied and damaged, which will damage the local ecological environment quality. With the country's strong advocacy of ecological civilization construction and high-quality development, green mines have become the theme of the future development of mines. During the process of mining in the mining area, evaluating and analyzing the changes in the ecological environment and predicting the future ecological environment quality of the mining area based on the existing conditions will help the mining area formulate regional ecological protection and reclamation design plans, rationally utilize mineral resources, and coordinate the mutual relationship between regional economy and ecological environment protection and governance. However, most of the existing research methods focus on the changes in the ecological environment quality in the past mining areas, and there are still obvious deficiencies in the prediction of the future ecological environment quality of the mining area. This limitation leads to the inability to adjust the mining area development strategy in a timely manner according to the future environmental change trend, thus affecting the sustainable development of the mining area. In addition, traditional methods mainly consider the impact of natural conditions on the ecological environment quality during the process of studying the ecological environment quality, and cannot comprehensively master the change law of the ecological environment quality. Therefore, there is an urgent need for a method that can simultaneously consider human activities and natural factors and accurately predict the future ecological environment quality of the mining area. Summary of the Invention

[0003] The purpose of the present invention is to solve the technical problems pointed out in the background art, and provide a method for predicting the ecological environment quality based on time-series remote sensing images and residual reconstruction. The climate factor regression unit uses the data subset only affected by climate factors to perform regression training in the raster map through the constructed regression function A to obtain the coefficients and adjustment terms of each climate factor. Then, the multiple regression residual analysis unit predicts the ecological environment quality A for the data subset jointly affected by climate factors and human activities according to the regression function A after regression training, and calculates the difference with the ecological environment quality data to obtain the residual. Then, the human activity factor regression unit uses the data subset jointly affected by climate factors and human activities to perform regression training of human activity factors in the raster map through the constructed regression function B to obtain the coefficients and adjustment terms of each human activity factor, and finally obtains the comprehensive ecological environment quality prediction function model.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] A method for predicting the ecological environment quality based on time-series remote sensing images and residual reconstruction, the method comprising:

[0006] S1. Obtain a long - time - series ecological environment quality sample data set within the research mining area that contains geographical coordinate information and records ecological environment quality data, and divide the ecological environment quality sample data set into a data subset only affected by climate factors and a data subset affected by both climate factors and human activities;

[0007] S2. Construct a multiple regression model including a raster map. The multiple regression model includes a climate factor regression unit, a multiple regression residual analysis unit, and a human activity factor regression unit. The climate factor regression unit constructs a regression function A with the ecological environment quality A as the dependent variable and several climate factor data as the independent variables, and uses the data subset only affected by climate factors to perform regression training of the regression function A in the raster map to obtain the coefficients and adjustment terms of the climate factors;

[0008] The multiple regression residual analysis unit predicts the ecological environment quality A for the data subset affected by both climate factors and human activities according to the regression function A after regression training, and obtains the residual through subtraction calculation with the ecological environment quality data;

[0009] The human activity factor regression unit constructs a regression function B with the residual data as the ecological environment quality B and as the dependent variable and several human activity factor data as the independent variables, and uses the human activity data in the data subset affected by both climate factors and human activities to perform regression training of the regression function B in the raster map to obtain the coefficients and adjustment terms of the human activity factors;

[0010] The multiple regression model adds the trained regression function A and regression function B in the raster map to form a comprehensive ecological environment quality prediction function model;

[0011] S3. Collect the climate factor data and human activity data within the research mining area and input them into the multiple regression model. The multiple regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction result in the raster map.

[0012] To better implement the present invention, the climate factors in the climate factor regression unit include average precipitation and average temperature , and the data subset only affected by climate factors contains at least the data corresponding to average precipitation , average temperature . The expression of the regression function A is as follows:

[0013] , where represents the ecological environment quality A, , are respectively , 's coefficients, is the adjustment term for the regression function A;

[0014] In the human activity factor regression unit, the human activity factors include the mining intensity MI and the mining area protection intensity MP of the mining area. The data subset affected by both climate factors and human activities contains at least the average precipitation , the average temperature , the data corresponding to the mining intensity MI and the mining area protection intensity MP of the mining area, and / or the data required for calculation. The expression of the regression function B is as follows:

[0015] , where represents the ecological environment quality B, , are the coefficients of MI and MP respectively, is the adjustment term for the regression function B;

[0016] In the comprehensive ecological environment quality prediction function model, the expression of the prediction function is as follows:

[0017] .

[0018] Preferably, the multiple regression model makes predictions according to the time scale, and the time scale is annual, quarterly or monthly. The data corresponding to the climate factors in the climate factor regression unit and the data corresponding to the human activity factors in the human activity factor regression unit are statistically analyzed according to the same time scale. The average precipitation and the average temperature in the climate factor regression unit, as well as the data corresponding to the mining intensity MI and the mining area protection intensity MP in the human activity factor regression unit, are all statistically analyzed annually.

[0019] Preferably, the mining intensity MI of the mining area is the settlement amount within the time scale, and the settlement amount is obtained by the following method:

[0020] Obtain multiple radar images within the time scale of the research mining area and calculate the deformation phase , and then calculate the settlement amount according to the following formula :

[0021] , where is the wavelength of the radar; represents the settlement amount, and the settlement amount is the vertical settlement amount; is the incident angle of the SAR satellite.

[0022] Preferably, the calculation method of the deformation phase is as follows:

[0023] Obtain the DEM image within the time scale of the research mining area and calculate the phase difference caused by terrain undulation ; Take one of the multi - radar images within the time scale of the research mining area as the main image, and register and perform interference processing on the remaining radar images. The differential interference phase expression of the interference processing is as follows:

[0024] , where represents the differential interference phase, represents the phase difference caused by the satellite orbit, represents the phase difference caused by atmospheric delay, represents the phase difference caused by noise;

[0025] Based on the above formula, the deformation phase is calculated.

[0026] Preferably, the mining area protection intensity MP is the improvement amount of NDVI within the time scale, and is obtained according to the following method:

[0027] B1. Obtain the remote sensing image data within the time scale of the research mining area and calculate the NDVI data. The calculation method of the NDVI data is as follows:

[0028] , where NIR represents the reflectance of the near - infrared band in the remote sensing image data, and R represents the reflectance of the red - light band in the remote sensing image data;

[0029] B2. The calculation method of the NDVI improvement amount of the time - scale sequence relative to the previous time - scale sequence is as follows:

[0030] ;

[0031] ;

[0032] where represents the NDVI deviation amount of the time - scale sequence tk, represents the time - scale sequence of the NDVI data, represents the time - scale sequence relative to the time - scale sequence of the NDVI improvement amount, used to characterize the mining area protection intensity MP, is the parameter fitted based on the historical data of the data subset only affected by climate factors.

[0033] Preferably, the parameter is obtained as follows:

[0034] Based on the ecological environment quality sample data set, use the vegetation disturbance - related algorithm to determine the time of the first vegetation disturbance in the research mining area, denoted as tc at a time earlier than t c of time t 1 t 2 t 3 t c-1 taking times t, t, t, ……, t earlier than t as independent variables and the NDVI data at the time t to be calculated as the dependent variable to construct a function model:

[0035] , and are all model parameters, is the slope in the function model and is used as the parameter for the change in NDVI under natural conditions within the time scale; t represents time, represents the NDVI data at time t;

[0036] Based on the above formula, the parameter is calculated.

[0037] An ecological environment quality prediction model based on time - series remote sensing images and residual reconstruction, including a sample data construction module, a data acquisition module, and a multiple regression model. The sample data construction module is used to obtain an ecological environment quality sample data set with long - time series, containing geographic coordinate information and recording ecological environment quality data within the research mining area, and divide the ecological environment quality sample data set into a data subset only affected by climate factors and a data subset affected by both climate factors and human activities. The multiple regression model internally contains a raster map. The multiple regression model includes a climate factor regression unit, a multiple regression residual analysis unit, and a human activity factor regression unit. The climate factor regression unit constructs a regression function A with the ecological environment quality A as the dependent variable and several climate factor data as the independent variables, and uses the data subset only affected by climate factors to perform regression training of the regression function A in the raster map to obtain the coefficients and adjustment terms of the climate factors. The multiple regression residual analysis unit is used to predict the ecological environment quality A for the data subset affected by both climate factors and human activities according to the regression function A after regression training, and calculate the residual by taking the difference from the ecological environment quality data. The human activity factor regression unit constructs a regression function B with the residual data as the ecological environment quality B as the dependent variable and several human activity factor data as the independent variables, and uses the human activity data in the data subset affected by both climate factors and human activities to perform regression training of the regression function B in the raster map to obtain the coefficients and adjustment terms of the human activity factors;

[0038] The multiple regression model adds the trained regression function A and regression function B in the raster map to form a comprehensive ecological environment quality prediction function model; the data acquisition module is used to collect climate factor data and human activity data in the research mining area and input them into the multiple regression model, and the multiple regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction result in the raster map.

[0039] Preferably, the climate factors in the climate factor regression unit include average precipitation and average temperature , and the data subset affected only by climate factors contains at least the data corresponding to average precipitation , average temperature . The expression of regression function A is as follows:

[0040] , where represents the ecological environment quality A, , are respectively , 's coefficients, is the adjustment term of regression function A;

[0041] The human activity factors in the human activity factor regression unit include the mining intensity MI and the mining protection intensity MP of the mining area. The data subset affected by both climate factors and human activities contains at least the data corresponding to average precipitation , average temperature , the mining intensity MI and the mining protection intensity MP of the mining area, or / and the data required for calculation. The expression of regression function B is as follows:

[0042] , where represents the ecological environment quality B, , are respectively the coefficients of MI and MP, is the adjustment term of regression function B;

[0043] The expression of the prediction function in the comprehensive ecological environment quality prediction function model is as follows:

[0044] .

[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0046] (1) The climate factor regression unit of the present invention uses the data subset affected only by climate factors to obtain the coefficients and adjustment terms of each climate factor through regression training in the raster map by constructing the regression function A. Then, the multiple regression residual analysis unit predicts the ecological environment quality A for the data subset affected by both climate factors and human activities according to the regression function A after regression training and calculates the difference with the ecological environment quality data to obtain the residual. Then, the human activity factor regression unit uses the data subset affected by both climate factors and human activities to perform regression training of human activity factors in the raster map by constructing the regression function B to obtain the coefficients and adjustment terms of each human activity factor, and finally obtains the comprehensive ecological environment quality prediction function model. The present invention uses the ecological environment quality sample data set to reconstruct the residual and trains the regression function A and the regression function B respectively, constructs a comprehensive ecological environment quality prediction function model considering natural factors and human activities, and solves the limitations of traditional methods in predicting the ecological environment quality of future mining areas.

[0047] (2) The present invention realizes the analysis and prediction of the ecological environment changes in the mining area by integrating multi-source long-time series remote sensing data, such as climate, rainfall, normalized difference vegetation index NDVI, and ecological environment quality data, and combining with interferometric synthetic aperture radar (InSAR) technology. The present invention not only improves the ability to capture the environmental change trend, but also explores the interactive influence between human activities and natural factors, providing comprehensive and accurate data support and decision-making basis for formulating scientific and reasonable mining area ecological protection and reclamation plans, optimizing mineral resource development strategies, and promoting the coordinated development of regional economy and ecological environment protection. Brief Description of the Drawings

[0048] Figure 1 is a schematic flow chart of the method for predicting the ecological environment quality of the present invention;

[0049] Figure 2 is a schematic principle diagram of residual reconstruction in Shendong Mining Area in the embodiment;

[0050] Figure 3 is a schematic principle diagram of extracting the ground settlement amount in Shendong Mining Area in the embodiment;

[0051] Figure 4 is a schematic principle diagram of calculating the mining area protection intensity in Shendong Mining Area in the embodiment. Detailed Embodiment

[0052] The present invention will be further described in detail below with reference to the embodiments: Embodiment

[0053] As Figure 1 shown, a method for predicting the ecological environment quality based on time-series remote sensing images and residual reconstruction, the method includes:

[0054] S1. Obtain a long - time - series ecological environment quality sample data set within the research mining area that contains geographical coordinate information and records ecological environment quality data, and divide the ecological environment quality sample data set into a data subset only affected by climate factors and a data subset affected by both climate factors and human activities.

[0055] In this embodiment, the Shendong Mining Area is selected as the research area in the embodiment of the present invention for illustration. The average annual precipitation, average annual temperature, and ecological environment quality data within the scope of the Shendong Mining Area from 1993 to 2023 in the past 30 years are obtained. Taking climate factors such as the average annual precipitation and average annual temperature before mining activities in the Shendong Mining Area as independent variables, and taking the ecological environment quality data as the dependent variable in this embodiment, a multiple - regression residual analysis method is applied to establish the relationship between climate factors and ecological quality. The specific method is as follows:

[0056] a. Data collection: Select the Shendong Mining Area as the research area, and collect the average annual precipitation, average annual temperature data, and ecological environment quality RSEI and other data of the Shendong Mining Area for 30 years from 1994 to 2023; these data can be obtained through channels such as remote - sensing satellites and meteorological stations.

[0057] b. To ensure the consistency and accuracy of the data, the collected data needs to be pre - processed; including operations such as data cleaning, format conversion, and resampling. The average annual precipitation, average annual temperature, and ecological environment quality RSEI and other data are screened and resampled to the same grid resolution (i.e., correspondingly sampled to a multiple - regression model including a grid map) for subsequent analysis and modeling.

[0058] S2. Construct a multiple - regression model including a grid map. The multiple - regression model includes a climate - factor regression unit, a multiple - regression residual - analysis unit, and a human - activity - factor regression unit. The climate - factor regression unit constructs a regression function A with the ecological environment quality A as the dependent variable and several climate - factor data as independent variables, and uses the data subset only affected by climate factors to perform regression training of the regression function A in the grid map and obtain the coefficients and adjustment terms of the climate factors.

[0059] In this embodiment, the climate factors in the climate - factor regression unit include average precipitation and average temperature (In this embodiment, the climate factors in the climate - factor regression unit select average precipitation and average temperature as two climate factors and are the technical preference of the present invention; in special environments, of course, other climate factors can also be added, such as topography. Average precipitation and average temperature The same time scale needs to be selected. The time scale can be annual, quarterly, or monthly. For example, if the time scale is selected as one year, the average precipitation and the average temperature will be the annual average precipitation and the annual average temperature. The data of each unit in the multiple regression model also follows the same time scale (for example, the time scale is all annual). The data subset that is only affected by climate factors should at least include the average precipitation , the average temperature corresponding data. The expression of regression function A is as follows:

[0060] , where represents the ecological environment quality A, , are respectively , 's coefficients, is the adjustment term of regression function A. If the time scale is selected as annual (i.e., one year), the average precipitation , the average temperature both correspond to the annual data, and the climate factor regression unit processes the data with the annual time scale. Taking the time scale selected as annual as an example, the average precipitation , the average temperature correspond to the annual average precipitation and the annual average temperature. The data subset that is only affected by climate factors is the annual average temperature, annual average precipitation, and ecological environment quality (the ecological environment quality is the quality data obtained by expert evaluation) within the study area range obtained from the time period of the study mining area that is only affected by climate factors; resample the data such as the annual average temperature, annual average precipitation, and ecological environment quality to the same grid resolution. Taking climate factors such as the annual average temperature and annual average precipitation as independent variables and RSEI (i.e., the ecological environment quality) as the dependent variable, use the multiple residual analysis method to establish a regression model between climate factors and the ecological environment quality RSEI pixel by pixel and calculate each coefficient of the above formula in the model.

[0061] In the example of taking Shendong Mining Area as the study mining area in this embodiment, the climate factor regression unit constructs regression function A with the ecological environment quality A as the dependent variable and several climate factor data as the independent variables, and uses the data subset that is only affected by climate factors to perform regression training of regression function A pixel by pixel (or grid by grid) in the grid map to obtain the coefficients and adjustment terms of the climate factors.

[0062] In the example of taking Shendong Mining Area as the study mining area in this embodiment, calculate the predicted value of the ecological environment quality that is only affected by climate factors after mining activities according to the fitting relationship between climate factors and the ecological environment quality obtained by the multiple regression residual analysis method (i.e., the residual calculation expression), such as Figure 2As shown in the figure, the difference between the actual value and the predicted value of the ecological environment quality is calculated based on the actual ecological quality data in the Shendong mining area as the residual, and the impact of human activities on the ecological environment quality of the study area is separated. The multiple regression residual analysis unit predicts the ecological environment quality A of the data subset affected by climate factors and human activities according to the regression function A after regression training, and calculates the residual by subtracting the ecological environment quality data. The residual calculation expression is as follows:

[0063] ;

[0064] where represents the residual of the ecological environment quality (reflecting the impact degree of human activities on the ecological environment quality of the study area, a positive residual value indicates that the impact of human activities on the ecological environment quality is positive, and a negative residual value indicates that the impact of human activities on the ecological environment quality is negative), represents the actual value of the ecological environment quality (that is, the ecological environment quality data stored in the data subset affected by climate factors and human activities), represents the predicted value of the ecological environment quality only considering climate factors (the predicted value reflects the natural change trend of the ecological environment quality in the mining area without the interference of human activities).

[0065] The human activity factor regression unit uses the residual data as the ecological environment quality B and as the dependent variable, constructs the regression function B with several human activity factor data as the independent variables, and conducts the regression training of the regression function B using the human activity data in the data subset affected by climate factors and human activities in the raster map to obtain the coefficients and adjustment terms of the human activity factors.

[0066] In this embodiment, the human activity factors in the human activity factor regression unit include the mining intensity MI and the mining protection intensity MP of the mining area (in this embodiment, the human activity factors in the human activity factor regression unit select two human activity factors, the mining intensity MI and the mining protection intensity MP, as the technical preference of the present invention. The time scale of the human activity factor data is consistent with the time scale of the climate factor data. If the time scale of the climate factor data (including the average precipitation and the average temperature ) is selected as one year (i.e., annual), then the time scale of the human activity factor data is also selected as one year. The data subset affected by climate factors and human activities contains at least the average precipitation , the average temperature , the data corresponding to the mining intensity MI and the mining protection intensity MP of the mining area and / or the data required for calculation. The expression of the regression function B is as follows:

[0067] , where Denote the ecological environment quality as B, and are the coefficients of MI and MP respectively, is the adjustment term of the regression function B.

[0068] The multiple regression model adds the trained regression function A and the regression function B in the raster map to form a comprehensive ecological environment quality prediction function model.

[0069] In this embodiment, the prediction function expression in the comprehensive ecological environment quality prediction function model is as follows:

[0070] .

[0071] S3. Collect the climate factor data and human activity data in the research mining area and input them into the multiple regression model. The multiple regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction result in the raster map.

[0072] Preferably, the multiple regression model makes predictions according to the time scale, and the time scale is annual, quarterly or monthly. The data corresponding to the climate factors in the climate factor regression unit and the data corresponding to the human activity factors in the human activity factor regression unit are statistically analyzed according to the same time scale. In this embodiment, the average precipitation and the average temperature in the climate factor regression unit, as well as the data corresponding to the mining intensity MI and the mining protection intensity MP in the human activity factor regression unit, are all statistically analyzed annually (the same time scale).

[0073] In some embodiments, the mining intensity MI in the mining area is the settlement amount within the time scale, and the settlement amount is obtained by the following method:

[0074] Obtain multiple radar images within the time scale of the research mining area to calculate the deformation phase , and then calculate the settlement amount according to the following formula :

[0075] , where is the wavelength of the radar; denotes the settlement amount, and the settlement amount is the vertical settlement amount; is the incident angle of the SAR satellite.

[0076] If the time scale is selected as annual (i.e., one year), then the deformation phase and the settlement amount both correspond to the annual data, and the human activity factor regression unit processes the data with the annual time scale.

[0077] In the example where the Shendong mining area is taken as the research mining area in this embodiment, as Figure 3 shown, the method for obtaining the mining intensity MI of the mining area is as follows:

[0078] a. Obtain multiple radar images within the Shendong mining area and perform InSAR processing to obtain the long-term deformation monitoring results of the research area;

[0079] b. Obtain the DEM image within the Shendong mining area to eliminate the influence of terrain undulation on the deformation monitoring results.

[0080] c. Select one image as the main image from N + 1 images, and perform registration and interference processing with the remaining auxiliary images to obtain N pairs of interferometric images. The differential interferometric phase can be expressed as:

[0081] ; the deformation phase can be further expressed as: ; the mining intensity of the Shendong mining area can be obtained based on the vertical settlement amount.

[0082] In some embodiments, the calculation method of the deformation phase is as follows:

[0083] Obtain the DEM image within the time scale of the research mining area and calculate the phase difference caused by terrain undulation ; select one radar image from multiple radar images within the time scale of the research mining area as the main image and perform registration and interference processing on the remaining radar images. The expression of the differential interferometric phase of the interference processing is as follows:

[0084] , where represents the differential interferometric phase, represents the phase difference caused by the satellite orbit, represents the phase difference caused by atmospheric delay, represents the phase difference caused by noise;

[0085] Based on the above formula, the deformation phase is calculated.

[0086] If the time scale is selected as annual (i.e., one year), the mining intensity MI is defined as the number of millimeters of settlement within the pixel range in one year. MI is the number of millimeters of settlement within the pixel range in one year, and is calculated using the SBAS-InSAR technology through radar image data and DEM data of the research area. The specific calculation process is as follows:

[0087] a. Obtain multiple radar images within the research area and perform InSAR processing to obtain the long-term deformation monitoring results of the research area;

[0088] b. Obtain the DEM image within the study area to eliminate the influence of terrain undulation on the deformation monitoring results;

[0089] c. Select one image as the main image from N + 1 images, and perform registration and interference processing with the remaining auxiliary images to obtain N pairs of interferometric images. The differential interferometric phase can be expressed as:

[0090] ;

[0091] where represents the differential interferometric phase, represents the phase difference caused by the satellite orbit, represents the phase difference caused by the atmospheric delay, represents the phase difference caused by the noise; It can be further expressed as:

[0092] ;

[0093] where is the wavelength of the radar; represents the settlement amount, and the settlement amount is the vertical settlement amount; is the incident angle of the SAR satellite, and the mining intensity within the study area can be calculated based on the vertical settlement amount.

[0094] In some embodiments, the mining area protection intensity MP is the improvement amount of NDVI within the time scale, and is obtained according to the following method:

[0095] B1. Obtain the remote sensing image data within the time scale of the study mining area and calculate the NDVI data. The calculation method of the NDVI data is as follows:

[0096] , where NIR represents the reflectance of the near-infrared band in the remote sensing image data, and R represents the reflectance of the red band in the remote sensing image data;

[0097] B2. The calculation method of the NDVI improvement amount of the time scale sequence relative to the previous time scale sequence is as follows:

[0098] ;

[0099] ;

[0100] where represents the NDVI deviation amount of the time scale sequence tk, represents the NDVI data of the time scale sequence , represents the time scale sequence Relative to the time scale sequence The improvement amount of NDVI is used to characterize the mining area protection intensity MP, which is a parameter obtained by fitting the historical data of the data subset affected only by climate factors. First, calculate , and then judge the relationship with zero. If , then take "0"; if , then take .

[0101] Among them, the method for obtaining the parameter is as follows:

[0102] Based on the ecological environment quality sample data set, use the vegetation disturbance related algorithm to judge the time of the first vegetation disturbance in the study mining area, denoted as t c . Take the time t c earlier than t 1 , t 2 , t 3 , ……, t c-1 as independent variables, and use the NDVI data at the time t to be calculated as the dependent variable to construct a function model:

[0103] , , are all model parameters, is the slope in the function model and is used as the parameter of the NDVI change amount under natural conditions within the time scale; t represents time, represents the NDVI data at time t;

[0104] Based on the above formula, calculate the parameter .

[0105] In this example, taking the Shendong mining area as the research mining area, the method for obtaining the mining area exploitation intensity MI is as follows:

[0106] a. Data collection: Obtain the long-term remote sensing images (including data such as Landsat5, Landsat7, Landsat8, etc.) of the Shendong mining area from 1994 to 2023 for 30 years, extract the long-term NDVI data and collect the corresponding year information. The NDVI calculation formula is as follows:

[0107] ; where NIR represents the reflectance of the near-infrared band and R represents the reflectance of the red band;

[0108] b. Use the vegetation disturbance related algorithm to judge the time of the first vegetation disturbance in the mining area range, denoted as t c , for exampleFigure 4 As shown, using the year information t c earlier than t 1 , t 2 , t 3 , ……, t c -1 as the independent variable and the NDVI data of the corresponding year as the dependent variable to establish a linear model: ;

[0109] Where The slope in the regression analysis represents the annual change value of NDVI under natural conditions and applies this change value to each subsequent year.

[0110] c. Calculation of NDVI improvement amount. For the year t c and the years after it k , calculate the improvement amount of NDVI. The calculation process is as follows: ;

[0111] ;

[0112] The protection intensity MP of the mining area can be characterized according to the improvement amount of NDVI.

[0113] In this example, taking the Shendong mining area as the research mining area, an ecological environment quality prediction model considering both natural conditions and human activities in the Shendong mining area is simulated through climate data, NDVI data, InSAR images, etc. before 2022. 1000 points are selected within the mining area as the verification sample set, and the ecological environment quality in 2023 is predicted through the relevant data in 2023 and the simulation accuracy is evaluated with the known ecological environment quality in 2023. The coefficient of determination (R 2 ) is used as the evaluation index of the interpretation rate to verify the simulation results. R 2 is the fitting degree between the predicted value and the observed value. The closer R 2 is to 1, the higher the prediction accuracy of the model. The calculation expression of the coefficient of determination R 2 is as follows:

[0114] ;

[0115] In the formula, n represents the number of samples, y i represents the actual observed value, represents the predicted value, represents the average value of the actual observed values, and the larger R 2 , the higher the simulation accuracy.

[0116] In this example, the specific steps of model stability analysis are as follows:

[0117] a. Data partitioning randomly divides the entire dataset (including climate factor data, RSEI data, MI, and MP data) into a training set and a validation set. 80% of the data is used for training, and the remaining is used for validation. b. Cross-validation subset partitioning: In the training set, subsets for k-fold cross-validation are further partitioned. In this embodiment, 5-fold cross-validation is adopted, and the training set is evenly divided into 5 subsets. Each time, one subset is set aside as the internal validation set, and the remaining 4 subsets are used as the internal training set. c. Model training and internal validation (cross-validation): For each fold of cross-validation, the model (including the regression model of climate factors and RSEI and the multiple regression model of human activity impacts) is retrained using the internal training set. d. Internal validation: The performance of the model in each fold of cross-validation is evaluated using the internal validation set. Error metrics such as mean squared error (MSE), root mean squared error (RMSE), or R² score are recorded for each validation. The average value and standard deviation of the error metrics in all folds of cross-validation are calculated. These statistics can reflect the performance stability and consistency of the model on different subsets. The changes in model parameters (such as regression coefficients) in each fold of cross-validation are also checked. If the parameter values are relatively stable, it indicates that the model is not highly dependent on the data and has good generalization ability. e. External validation: After completing the cross-validation, the final model is retrained using the entire training set (excluding the subsets used for cross-validation partitioning). The validation set is used to perform external validation on the final model to evaluate its performance in practical applications.

[0118] An ecological environment quality prediction model based on time-series remote sensing images and residual reconstruction, including a sample data construction module, a data acquisition module, and a multiple regression model. The sample data construction module is used to obtain a long-time series ecological environment quality sample data set within the research mining area that includes geographical coordinate information and records ecological environment quality data, and divide the ecological environment quality sample data set into a data subset only affected by climate factors and a data subset affected by both climate factors and human activities. The multiple regression model internally contains a raster map. The multiple regression model includes a climate factor regression unit, a multiple regression residual analysis unit, and a human activity factor regression unit. The climate factor regression unit constructs a regression function A with the ecological environment quality A as the dependent variable and several climate factor data as the independent variables, and uses the data subset only affected by climate factors to perform regression training of the regression function A in the raster map to obtain the coefficients and adjustment terms of the climate factors. The multiple regression residual analysis unit is used to predict the ecological environment quality A for the data subset affected by both climate factors and human activities according to the regression function A after regression training, and calculate the difference from the ecological environment quality data to obtain the residual. The human activity factor regression unit constructs a regression function B with the residual data as the ecological environment quality B and as the dependent variable and several human activity factor data as the independent variables, and uses the human activity data in the data subset affected by both climate factors and human activities to perform regression training of the regression function B in the raster map to obtain the coefficients and adjustment terms of the human activity factors.

[0119] The multiple regression model adds the trained regression function A and regression function B in the raster map to form a comprehensive ecological environment quality prediction function model. The data acquisition module is used to collect climate factor data and human activity data within the research mining area and input them into the multiple regression model. The multiple regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction result in the raster map.

[0120] In the climate factor regression unit preferably of the present invention, the climate factors include average precipitation and average temperature , and the data subset only affected by climate factors contains at least the data corresponding to average precipitation and average temperature . The expression of the regression function A is as follows:

[0121] , where represents the ecological environment quality A, , are respectively , the coefficients of is the adjustment term of the regression function A;

[0122] In the preferred human activity factor regression unit of the present invention, the human activity factors include the mining intensity MI and the mining area protection intensity MP. The data subset affected by both climate factors and human activities at least includes the average precipitation , the average temperature , the data corresponding to the mining intensity MI and the mining area protection intensity MP, and / or the data required for calculation. The expression of the regression function B is as follows:

[0123] , where represents the ecological environment quality B, , are the coefficients of MI and MP respectively, is the adjustment term of the regression function B;

[0124] The expression of the prediction function in the comprehensive ecological environment quality prediction function model is as follows:

[0125] .

[0126] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction, characterized by: The methods include: S1. Obtain a long-term ecological environment quality sample data set in the research mining area that contains geographic coordinate information and records ecological environment quality data, and divide the ecological environment quality sample data set into a data subset affected only by climate factors and a data subset affected by both climate factors and human activities; S2. Construct a multiple regression model including a grid map. The multiple regression model includes a climate factor regression unit, a multiple regression residual analysis unit, and a human activity factor regression unit. The climate factor regression unit uses the ecological environment quality A as the dependent variable and several climate factor data as independent variables to construct a regression function A. The regression function A is trained on the grid map using only a subset of data affected by climate factors to obtain the coefficient and adjustment term of the climate factor. The multivariate regression residual analysis unit predicts the ecological environment quality A of the data subset affected by climate factors and human activities according to the regression function A after regression training, and obtains the residual by subtracting it from the actual value of the ecological environment quality; The human activity factor regression unit uses the residual data as the ecological environment quality B and as the dependent variable, and uses several human activity factor data as independent variables to construct a regression function B, and uses the human activity data in the data subset affected by climate factors and human activities to perform regression training of the regression function B in the grid map and obtain the coefficient and adjustment term of the human activity factor; The multivariate regression model adds the trained regression function A and regression function B in the grid map to form a comprehensive ecological environment quality prediction function model; S3. Collect climate factor data and human activity data in the research mining area and input them into the multivariate regression model. The multivariate regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction results in the raster map.

2. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 1 is characterized by: The climate factors in the climate factor regression unit include average precipitation and average temperature , only the data subset affected by climate factors contains at least the average precipitation , average temperature The corresponding data, the regression function A expression is as follows: ,in A represents the ecological environment quality, , They are , The coefficient of is the adjustment term of regression function A; The human activity factors in the human activity factor regression unit include mining intensity MI and mining protection intensity MP. The data subset affected by climate factors and human activities at least includes average precipitation. , average temperature , the data corresponding to the mining intensity MI and the mining area protection intensity MP or / and the data required for calculation, the regression function B is expressed as follows: ,in Indicates the ecological environment quality B, , are the coefficients of MI and MP respectively, The adjustment term of regression function B; The prediction function expression in the comprehensive ecological environment quality prediction function model is as follows: 。 3. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 2 is characterized by: The multivariate regression model performs prediction according to a time scale, which is annual, quarterly or monthly. The data corresponding to the climate factor in the climate factor regression unit and the data corresponding to the human activity factor in the human activity factor regression unit are statistically analyzed according to the same time scale.

4. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 3 is characterized by: The mining intensity MI of the mining area is the settlement amount within the time scale, and the settlement amount is obtained according to the following method: Obtain multiple radar images within the time scale of the research mining area to calculate the deformation phase , and then calculate the settlement according to the following formula : ,in is the wavelength of the radar; Indicates the amount of settlement, which is the vertical settlement; is the incident angle of the SAR satellite.

5. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 4 is characterized by: Deformation Phase The calculation method is as follows: Obtain DEM images within the time scale of the research mining area and calculate the phase difference caused by terrain undulation ; From the multiple radar images within the time scale of the research mining area, one radar image is taken as the main image and the remaining radar images are registered and interferometrically processed. The differential interferometric phase expression of the interferometric processing is as follows: ,in represents the differential interference phase, represents the phase difference caused by the satellite orbit, represents the phase difference caused by atmospheric delay, Indicates that noise causes phase difference; The deformation phase is calculated based on the above formula .

6. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 2 is characterized by: The mining area protection intensity MP is the improvement of NDVI within the time scale, which is obtained as follows: B1. Obtain remote sensing image data within the time scale of the research mining area and calculate the NDVI data. The calculation method of NDVI data is as follows: , where NIR represents the reflectivity of the near infrared band in the remote sensing image data, and R represents the reflectivity of the red light band in the remote sensing image data; B2. Time scale series Relative to the previous time scale series The NDVI improvement is calculated as follows: ； ； in represents the NDVI deviation of the time scale series tk, Represents a time scale series NDVI data, Represents a time scale series Relative to the time scale series The NDVI improvement is used to characterize the protection intensity MP of the mining area. are the parameters fitted based on the historical data of the subset of data affected only by climate factors.

7. The method for predicting ecological environment quality based on time series remote sensing images and residual reconstruction according to claim 6 is characterized by: parameter The method of obtaining is as follows: Based on the ecological environment quality sample data set, the time when the first vegetation disturbance occurred in the research mining area was determined by the vegetation disturbance correlation algorithm as tc. The time t1, t2, t3, ..., tc-1 earlier than tc was used as the independent variable, and the NDVI data at the time t to be calculated was used as the dependent variable to construct the function model: , , are model parameters, is the slope in the function model and serves as a parameter for the change in NDVI under natural conditions within a time scale; t represents time, Represents the NDVI data at time t; The parameters are calculated based on the above formula .

8. An ecological environment quality prediction system based on time series remote sensing images and residual reconstruction, characterized by: The invention comprises a sample data construction module, a data collection module and a multiple regression model. The sample data construction module is used to obtain a long-time series of an ecological environment quality sample data set in a research mining area that contains geographic coordinate information and records ecological environment quality data, and divides the ecological environment quality sample data set into a data subset affected only by climate factors and a data subset affected by both climate factors and human activities. The multiple regression model contains a grid map. The multiple regression model comprises a climate factor regression unit, a multiple regression residual analysis unit and a human activity factor regression unit. The climate factor regression unit takes the ecological environment quality A as the dependent variable and several climate factor data as the independent variables to construct a regression function A and uses the data affected only by climate factors to construct a regression function A. The data subset is subjected to regression training of regression function A in a grid map and the coefficient and adjustment item of the climate factor are obtained; the multivariate regression residual analysis unit is used to predict the ecological environment quality A of the data subset affected by climate factors and human activities according to the regression function A after regression training, and obtain the residual by performing difference calculation with the ecological environment quality data; the human activity factor regression unit constructs the regression function B with the residual data as the ecological environment quality B and as the dependent variable and with a number of human activity factor data as the independent variable, and uses the human activity data in the data subset affected by climate factors and human activities to perform regression training of the regression function B in a grid map and obtain the coefficient and adjustment item of the human activity factor; The multivariate regression model adds the trained regression function A and regression function B in the grid map to form a comprehensive ecological environment quality prediction function model; the data acquisition module is used to collect climate factor data and human activity data in the research mining area and input them into the multivariate regression model, and the multivariate regression model calculates and processes according to the comprehensive ecological environment quality prediction function model and outputs the ecological environment quality prediction result in the grid map.

9. The ecological environment quality prediction system based on time series remote sensing images and residual reconstruction according to claim 8 is characterized by: The climate factors in the climate factor regression unit include average precipitation and average temperature , only the data subset affected by climate factors contains at least the average precipitation , average temperature The corresponding data, the regression function A expression is as follows: ,in A represents the ecological environment quality, , They are , The coefficient of is the adjustment term of regression function A; The human activity factors in the human activity factor regression unit include mining intensity MI and mining protection intensity MP. The data subset affected by climate factors and human activities at least includes average precipitation. , average temperature , the data corresponding to the mining intensity MI and the mining area protection intensity MP or / and the data required for calculation, the regression function B is expressed as follows: ,in Indicates the ecological environment quality B, , are the coefficients of MI and MP respectively, is the adjustment term of regression function B; The prediction function expression in the comprehensive ecological environment quality prediction function model is as follows: 。

Images