A method for identifying ecological damage areas in an open coal mine area based on time-series remote sensing images

Abstract

The application discloses a kind of open coal mine area ecological damage area identification method based on time sequence remote sensing image, belong to open coal mine area ecological damage area identification technical field.First, the natural environment data of study area is collected, the time sequence dataset of vegetation index and meteorological parameter is established, the correlation coefficient influence between preparation index and local meteorological parameter is determined, then the vegetation type is determined according to significant difference, preparation-meteorological correlation index is established, finally, according to the domain of the correlation index of different preparation type, natural vegetation and reconstructed vegetation are identified, and the classification of natural vegetation and reconstructed vegetation in remote sensing image is completed.The steps are simple, the analysis effect is good, the precision is high, and it has wide practicability.

Description

Technical Field

[0001] This invention relates to a method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery, belonging to the technical field of ecological damage zone identification in open-pit coal mines. Background Technology

[0002] To understand the extent of ecological damage in open-pit coal mines and effectively restore them, it is necessary to accurately determine the area, distribution, and type of ecologically damaged zones. This is also an important task for natural resources and environmental protection departments. Because the ecological damage in open-pit coal mines is extensive and changes rapidly, traditional field surveys are insufficient to meet the needs of dynamic and precise monitoring. Therefore, remote sensing technology is required to identify ecologically damaged zones.

[0003] Currently, several methods exist for identifying and classifying ecologically damaged areas. For example, patent CN202211170995.7 developed a post-earthquake vegetation restoration monitoring method. This method primarily involves selecting earthquake-damaged areas to generate remote sensing images of the earthquake region, establishing remote sensing images of the earthquake region in different seasons and years, and extracting areas of sparse vegetation caused by the earthquake. It then quantifies the recovery status of various vegetation types and the overall vegetation. Patent CN202111044342.X developed a dynamic assessment method for ecological environment damage in different land function zones. This method selects the study area and divides it into unit grids, and uses the AHP method to establish a sensitivity evaluation system for different land types to the surface deformation index, obtaining the sensitivity coefficients of different land types to ecological environment damage. Patent CN202111340720.9 developed an ecological risk assessment method based on topographic gradients. This invention is achieved through four steps: constructing a simulation system, determining abiotic environmental conditions, identifying risk sources, and classifying and evaluating risk receptors and indicators.

[0004] However, the aforementioned methods struggle to address issues such as spectral similarity and temporal variability between vegetation in ecologically damaged areas and naturally growing areas, thus hindering rapid identification and accurate classification of ecologically damaged areas. To address this, this invention proposes a deep learning-based remote sensing method for identifying ecologically damaged areas. This method automatically extracts features of ecologically damaged areas from remote sensing image data and fully utilizes temporal information from multi-temporal data to improve identification accuracy and robustness. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this paper proposes a method for identifying ecological damage areas in open-pit coal mines based on time-series remote sensing imagery. This method can automatically extract features of ecological damage areas from remote sensing imagery data and fully utilize time-series information by using multi-temporal data to identify ecological damage areas caused by open-pit coal mining and classify vegetation. This method meets the needs of monitoring and management for ecological restoration projects. The steps are simple, the identification accuracy is high, and the robustness is strong.

[0006] To address the aforementioned problems, this invention discloses a method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery, the steps of which are as follows:

[0007] S1. Obtain natural environmental information of the target area for each month over several years through a remote sensing platform, including rainfall, accumulated temperature, average temperature, sunshine days data, as well as normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and ratio vegetation image data (RVI). Then, through resampling, raster calculation, and layer overlay, the above data are converted into meteorological image data and vegetation image data with the same spatial resolution. Meteorological image data and vegetation image data of the same month of the same year are grouped together as a set of data. Then, all data groups are grouped by month to form 12 groups. Finally, meteorological image data and vegetation image data groups of the same month of different years are arranged together.

[0008] S2. Perform pixel-by-pixel analysis on meteorological image data and vegetation image data in all data groups. Use bivariate correlation analysis to calculate the Pearson correlation coefficient, Kendall's tau-b correlation coefficient, Spearman rank correlation coefficient and significance level of each pixel between meteorological image data and vegetation image data. Then, obtain the vegetation-meteorological correlation coefficient image by overlaying the meteorological image data and vegetation image data into layers.

[0009] S3. Classify and define the vegetation-meteorological correlation coefficient image pixel by pixel. Pixels with a significance level greater than 0.05 and NDVI less than 0.2 are defined as ecological damage areas, while pixels with a significance level less than 0.05 and NDVI greater than 0.2 are defined as natural vegetation areas. Then, normalize the vegetation-meteorological correlation coefficient and calculate the mean to establish a vegetation-meteorological correlation index image.

[0010] S4. Based on historical remote sensing images of the same area, manually select known vegetation types, randomly select 30-50 pixels in ArcMap as sample points, and statistically analyze the value range of the vegetation-meteorological correlation index for different vegetation types. Finally, based on the value range, identify the type of vegetation in the ecological damage area and the natural vegetation area represented by each pixel in the target area. A high degree of correlation in the value range of the vegetation-meteorological correlation index indicates that it is mainly driven by nature, while a low degree of correlation indicates that it is driven by human activities.

[0011] Furthermore, the specific method of step S1 is as follows: Using remote sensing platforms such as Google Earth Engine (GEE) and Geospatial Data Cloud, download basic data for the target area, including rainfall, accumulated temperature, average temperature, sunshine days, and the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Image Data (RVI). Existing meteorological and vegetation image data products can be used directly. For missing meteorological image data, spatial mapping is performed using spatial interpolation analysis methods in software such as GEE, ENVI, and ArcMap, based on data from Chinese meteorological stations. For missing vegetation image data, after downloading the original remote sensing images of the target area, the NDVI, EVI, and RVI of the study area are calculated using band calculations. Then, in platforms / software such as GEE, the meteorological and vegetation image data are converted into raster data with the same spatial resolution through resampling. For data from multiple periods within the same month, the average value is calculated using a band calculator as the data for that month.

[0012] Further, the specific method of step S2 is as follows: The meteorological and vegetation raster data established in S1 are used in data processing software / platform to perform correlation analysis on the values ​​of the same pixel in the meteorological and vegetation image data, calculating the Pearson correlation coefficient, Kendall's tau-b correlation coefficient, and Spearman rank correlation coefficient. This yields 12 groups of raster data, totaling 36 images, including precipitation-NDVI, precipitation-EVI, precipitation-RVI, accumulated temperature-NDVI, accumulated temperature-EVI, accumulated temperature-RVI, average temperature-NDVI, average temperature-EVI, average temperature-RVI, sunshine days-NDVI, sunshine days-EVI, and sunshine days-RVI, representing a total of 36 images with corresponding spatial distributions of correlation coefficients and their significance levels. These 12 groups of raster data are then overlaid and combined using the band combination function to obtain a single vegetation and meteorological correlation coefficient image synthesized from the 36 correlation coefficient bands. The data processing software / platform includes GEE, MatLab, and R language.

[0013] Furthermore, the normalization formula for the correlation coefficients of each band in the vegetation-meteorological correlation coefficient image is as follows:

[0014]

[0015] In the formula R ij Let R be the correlation coefficient of pixel j in the data with correlation coefficient i. imax R represents the maximum correlation coefficient in the i-th data set. imin The minimum value of the correlation coefficient in the data of correlation coefficient i;

[0016] However, in platforms / software such as GEE, the mean of the normalized correlation coefficient is calculated grid by grid using a raster calculator, and finally the vegetation-meteorological correlation index image is obtained.

[0017] Furthermore, the specific method of step S4 is as follows: a FeatureCollection is added to the map view area in the GEE platform to create sample points. After creating the sample points, 30-50 sample points of different types of vegetation are selected on the map and point data layers are established. The maximum, minimum, mean and standard layers of the vegetation-meteorological correlation index of the selected sample points are statistically analyzed. Thus, the vegetation-meteorological correlation index value range characteristics and discrimination criteria of different types of natural vegetation and reconstructed vegetation are determined. Finally, the corresponding natural vegetation and reconstructed vegetation types and their spatial distribution in the target area are screened according to the discrimination criteria.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1) By combining meteorological image data with vegetation image data, the accuracy of identification and classification of natural vegetation areas and ecological damage areas has been improved.

[0020] 2) It overcomes the problem of abnormal vegetation dynamics caused by the loss of remote sensing images or cloud cover in some years, and improves the robustness of the ecological damage area identification method. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery, as described in this invention.

[0022] Figure 2 This is a schematic diagram of the distribution of natural and reconstructed vegetation (willow and poplar) in the Ordos mining area in an embodiment of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and examples.

[0024] like Figure 1 As shown, this invention provides a method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery, using the mining area of ​​Ordos City, Inner Mongolia, as the research scope. The specific steps include:

[0025] S1. Acquire rainfall, accumulated temperature, average temperature, sunshine days data, normalized vegetation index (NDVI), enhanced vegetation index (EVI), and ratio vegetation image data (RVI) of the target area through remote sensing platforms such as Google Earth Engine and geospatial data cloud. Then, transform the above data into monthly meteorological image data and vegetation image data with the same spatial resolution through resampling, raster calculation, and layer overlay.

[0026] First, the TerraClimate data product was selected as the source of meteorological image data in Google Earth Engine. This data contains monthly meteorological parameters for the global period from 1958 to 2021, specifically cumulative rainfall, maximum temperature, and minimum temperature data. For vegetation image data, NDVI products from Landsat 8, Landsat 7, and Landsat 5 were selected, containing NDVI data for every 16 / 8 days (Landsat 5 every 16 days, Landsat 8 and Landsat 7 every 8 days) from 1989 to the present. The study period was selected as 2000-2020, and TerraClimate and Landsat series data for this time period were downloaded. The spatial resolution of the TerraClimate data product is 4638.3m, while the spatial resolution of the Landsat series data is 30m. Therefore, bilinear interpolation was used in ArcMap to resample the NDVI data to match the spatial resolution of the meteorological image data. Then, the NDVI data for the same month were averaged using a raster calculator to obtain the NDVI data layer for each month. Finally, the meteorological image data and vegetation image data for each month were overlaid and merged using the layer overlay function in ArcMap, thus obtaining the meteorological image data and vegetation image data layers for each month from 2000 to 2020.

[0027] S2. Perform pixel-by-pixel analysis on the meteorological image data and vegetation image data established in step S1. Use bivariate correlation analysis to calculate the Pearson correlation coefficient, Kendall'stau-b correlation coefficient, Spearman rank correlation coefficient and significance level of each pixel between the meteorological image data and the vegetation image data. Then, obtain the vegetation-meteorological correlation coefficient image by layer overlay.

[0028] In this example, the monthly meteorological image data and vegetation image data layers established in S1 are analyzed pixel-by-pixel using the GEE platform. Using bivariate correlation analysis, the Pearson correlation coefficient, Kendall's tau-b correlation coefficient, and Spearman rank correlation coefficient between the meteorological image data and vegetation image data for each pixel are calculated using the functions ee.Reducer.pearsonsCorrelation, ee.Reducer.kendallsCorrelation, and ee.Reducer.spearmansCorrelation. Furthermore, the significance test results of each correlation coefficient are saved as separate layers. This results in 12 groups of raster data layers, including rainfall-NDVI, rainfall-EVI, and rainfall-RVI, totaling 36 images showing the spatial distribution of correlation coefficients and 36 significance test results. Finally, in GEE, the layers of cumulative rainfall, maximum temperature, minimum temperature, and different types of remote sensing vegetation indices and their correlation coefficients are merged using band combination to obtain cumulative rainfall-vegetation, maximum temperature-vegetation, and minimum temperature-vegetation correlation coefficient images.

[0029] S3. Classify the correlation coefficient images established in step S2. Classify pixels with a significance level below 0.05 and an NDVI greater than 0.2 as natural vegetation areas, and classify pixels with a significance level greater than 0.05 and an NDVI greater than 0.2 as ecological damage areas. Then normalize the correlation coefficients and calculate the mean to finally establish the vegetation-meteorological correlation index image.

[0030] The images were filtered using the Map.image.lt function in GEE to select pixels with a significance level below 0.05, and these pixels were then removed using a masking function. The mean correlation coefficient of the remaining pixels was then calculated using the Map.image.mean function, thus obtaining the vegetation-meteorological correlation index image.

[0031] S4. Combine historical remote sensing images to manually select known vegetation types, select 30-50 sample points in ArcMap to count the value range of vegetation-meteorological correlation index for different vegetation types, identify each pixel representing natural vegetation or reconstructed vegetation based on the correlation index value range of different vegetation types, and classify them.

[0032] In this example, combining historical remote sensing imagery from Google Earth, ESAWorldCover 10m v100 land cover product data, and field survey results, two typical arid zone vegetation types, Pinus tabuliformis and Salix psammophila, were selected. Fifty sampling points for each type were chosen in GEE, and their vegetation-meteorological correlation index values ​​were statistically analyzed, as shown in Table 1.

[0033] Table 1. Range of meteorological correlation index values ​​for natural and reconstructed vegetation (Salix matsudana and Populus tomentosa).

[0034]

[0035] Finally, based on the meteorological correlation index ranges of different types of reconstructed vegetation, the spatial distribution of reconstructed and natural Pinus tabuliformis and Salix psammophila was determined, and mapping was performed in ArcMap through reclassification (e.g., Figure 2 As shown in the figure, this enables the identification and classification of natural vegetation and reconstructed vegetation within the study area.

Claims

1. A method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery, characterized in that... Includes the following steps: S1. Obtain natural environmental information of the target area for each month over several years through a remote sensing platform, including rainfall, accumulated temperature, average temperature, sunshine days data, as well as Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Image Data (RVI). Then, through resampling, raster calculation, and layer overlay, the above data are converted into meteorological image data and vegetation image data with the same spatial resolution. Meteorological image data and vegetation image data of the same year and month are grouped together as a set of data. Then, all data groups are grouped by month to form 12 groups. Finally, meteorological image data and vegetation image data groups of the same month in different years are arranged together. S2. Perform pixel-by-pixel analysis on meteorological image data and vegetation image data in all data groups. Use bivariate correlation analysis to calculate the Pearson correlation coefficient, Kendall's tau-b correlation coefficient, Spearman rank correlation coefficient and significance level of each pixel between meteorological image data and vegetation image data. Then, obtain the vegetation-meteorological correlation coefficient image by overlaying the meteorological image data and vegetation image data into layers. S3. Classify and define the vegetation-meteorological correlation coefficient image pixel by pixel. Pixels with a significance level greater than 0.05 and NDVI less than 0.2 are defined as ecological damage areas, while pixels with a significance level less than 0.05 and NDVI greater than 0.2 are defined as natural vegetation areas. Then, normalize the vegetation-meteorological correlation coefficient and calculate the mean to establish a vegetation-meteorological correlation index image. S4. Combine historical remote sensing images of the same area to manually select known vegetation types, randomly select 30-50 pixels in ArcMap as sample points to count the value range of vegetation-meteorological correlation index for different types of vegetation, and finally identify the type of vegetation in the ecological damage area and natural vegetation area represented by each pixel in the target area based on the value range.

2. The method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery according to claim 1, characterized in that, The specific method for step S1 is as follows: Using Google Earth Engine (GEE) and the Geospatial Data Cloud Remote Sensing Platform, download basic data for the target area, including rainfall, accumulated temperature, average temperature, sunshine days, Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Image Data (RVI). Existing meteorological and vegetation image data products are used directly. For missing meteorological image data, spatial mapping is performed using spatial interpolation analysis methods in GEE, ENVI, and ArcMap software based on data from Chinese meteorological stations. For missing vegetation image data, the NDVI, EVI, and RVI of the study area are obtained by downloading the original remote sensing images of the target area and calculating them using bands. Then, in the GEE platform / software, the meteorological and vegetation image data are converted into raster data with the same spatial resolution through resampling. For data from multiple periods within the same month, the average value is calculated using a band calculator as the data for that month.

3. The method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery according to claim 1, characterized in that, The specific method of step S2 is as follows: The meteorological and vegetation raster data established in S1 are used in data processing software / platform to perform correlation analysis on the values ​​of the same pixel in the meteorological and vegetation image data, calculating the Pearson correlation coefficient, Kendall's tau-b correlation coefficient, and Spearman rank correlation coefficient. This yields 12 groups of raster data, totaling 36 images, including precipitation-NDVI, precipitation-EVI, precipitation-RVI, accumulated temperature-NDVI, accumulated temperature-EVI, accumulated temperature-RVI, average temperature-NDVI, average temperature-EVI, average temperature-RVI, sunshine days-NDVI, sunshine days-EVI, and sunshine days-RVI, representing a total of 36 images with corresponding spatial distributions of correlation coefficients and their significance levels. These 12 groups of raster data are then combined using the band combination function to obtain a single vegetation and meteorological correlation coefficient image synthesized from the 36 correlation coefficient bands. The data processing software / platform includes GEE, MatLab, and R language.

4. The method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery according to claim 3, characterized in that, The formula for normalizing the correlation coefficients of each band in the vegetation-meteorological correlation coefficient image is as follows: （1）， In the formula Let i be the correlation coefficient of pixel j in the data. The maximum value of the correlation coefficient in the correlation coefficient i data. The minimum value of the correlation coefficient in the data of correlation coefficient i; However, in the GEE platform / software, the mean of the normalized correlation coefficient is calculated grid by grid using a raster calculator, and finally the vegetation-meteorological correlation index image is obtained.

5. The method for identifying ecological damage zones in open-pit coal mines based on time-series remote sensing imagery according to claim 1, characterized in that, The specific method of step S4 is as follows: a FeatureCollection is added to the map view area in the GEE platform to create sample points. After creating the sample points, 30-50 sample points of different types of vegetation are selected on the map and point data layers are established. The maximum, minimum, mean and standard layers of the vegetation-meteorological correlation index of the selected sample points are statistically analyzed. Thus, the vegetation-meteorological correlation index value range characteristics and discrimination criteria of different types of natural vegetation and reconstructed vegetation are determined. Finally, the corresponding natural vegetation and reconstructed vegetation types and their spatial distribution in the target area are screened according to the discrimination criteria.

Images