Regional ecological environment quality evaluation method based on neighborhood granular fuzzy c-means clustering

Abstract

The invention discloses a regional ecological environment quality evaluation method based on neighborhood granular fuzzy c-means clustering, and belongs to the technical field of electrical digital data processing, and the method comprises the steps: calculating and reconstructing a remote sensing time sequence MODIS data set based on a cloud platform, standardizing ecological factors, dividing ecological factor remote sensing images into n categories based on neighborhood granular fuzzy c-means clustering, and obtaining a comprehensive remote sensing ecological index PRSEI based on the MODIS by using an improved entropy weight method, and obtaining mining area ecological environment quality changes in year intervals according to the PRSEI of the research area. According to the method, the entropy weight method and the neighborhood fuzzy c-means clustering algorithm are effectively fused, all remote sensing pixel information of all ecological factors is reserved, the process is clear, and complex system problems such as remote sensing ecology and the like can be better analyzed and solved.

Description

Technical Field

[0001] The invention relates to a method for evaluating regional ecological environment quality based on neighborhood granular fuzzy C-means clustering, and belongs to the technical field of electrical digital data processing. Background Art

[0002] Remote sensing technology monitoring methods can achieve long-term, large-scale, real-time, and rapid dynamic monitoring of regional ecological environment, and provide information and decision-making services for regional environmental protection, restoration management, and sustainable development. Xu Hanqiu extracted four indicators of greenness, heat, humidity, and dryness based on Landsat remote sensing images, and constructed the traditional remote sensing ecological index RSEI using principal component analysis. The weight calculation uses principal component analysis (PCA), which can effectively extract the main information and reduce the data dimension. However, its orthogonal transformation characteristics based on continuous values ​​limit the in-depth understanding of the model's operating mechanism and the spatial distribution relationship of indicators.

[0003] Traditional ecological restoration strategies are mostly focused on the surface level, such as vegetation restoration and soil improvement. With the deepening of the understanding of ecosystems, restoration work has gradually shifted to vertical spatial extension, aiming to fully restore the multi-dimensional structure and function of ecosystems. At the same time, the evaluation of ecological environmental quality has also achieved a transformation from a single surface element to a three-dimensional spatial scale, in order to more comprehensively reflect the true status of the regional ecological environment. However, although the traditional remote sensing ecological index can reflect key ecological indicators such as vegetation coverage, surface humidity and soil dryness, it does not take into account the characteristics of atmospheric environmental changes, especially ignoring the important impact of fine particulate matter in mining areas on the atmospheric environment. In response to this limitation, some scholars have introduced PM 2.5 The remote sensing indicator DI (difference index, DI) of concentration is used to make up for the shortcomings of RSEI in atmospheric environment monitoring. However, the reflection, radiation and other signal values ​​represented by DI have large errors in accuracy and consistency with the actual ground monitoring data. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention proposes a method for evaluating the quality of regional ecological environment based on neighborhood granular fuzzy c-means clustering. The technical solution adopted by the present invention is: A method for evaluating regional ecological environment quality based on neighborhood granular fuzzy c-means clustering, characterized by comprising the following steps: Step 1: Calculate and reconstruct the remote sensing time series MODIS dataset and extract ecological factors; MODIS data products include MOD13A1, MOD11A2 and MOD09A1.

[0005] Step 2: Standardize the ecological factors; the MODIS data standardization process includes defining the cloud removal function and water mask function of the MODIS series satellite remote sensing image data.

[0006] Step 3: Divide the standardized remote sensing images of each ecological factor into n categories based on neighborhood granular fuzzy c-means clustering; including the following sub-steps: 3-1: Granulate the samples to form particle vectors, cluster the samples based on the distance between the particle vectors, and obtain the central membership of each sample point to all classes by optimizing the objective function to determine the class of the sample point and automatically divide the sample data; 3-2: Assign the membership function of each cluster to each sample, quantify the cluster to which the sample belongs based on the size of the membership, and further optimize the objective function to obtain the minimum value.

[0007] Step 4: The various ecological factors divided by the neighborhood granular fuzzy c-means clustering algorithm are used to obtain the MODIS-based comprehensive remote sensing ecological index PRSEI using the improved entropy weight method. The sub-steps are as follows: 4-1: Divide each ecological factor into different levels according to the knowledge granularity; 4-2: Calculate the knowledge granularity entropy based on the remote sensing information particles of different ecological factors; 4-3: Calculate the scale transformation knowledge granularity entropy based on the influence of the overall evaluation factors in different regions; 4-4: Calculate the weights of each ecological factor based on the scale transformation knowledge granularity entropy; 4-5: The PRSEI of the study area is obtained by multiplying the remote sensing images of each ecological factor with its weight and accumulating them;

[0008] Step 5: Based on the PRSEI of different years, the ecological environment quality of the study area is obtained according to the graded assignment method, and the change of ecological environment quality of the mining area in the year interval is obtained by the difference method. The graded assignment refers to dividing the PRSEI of a specific year into five levels: poor, relatively poor, medium, good, and excellent by the natural breakpoint method, and re-grading and assigning values ​​from 1 to 5; the difference method refers to subtracting the PRSEI grade values ​​of the starting year, and dividing the ecological quality changes into five levels: significantly worse, worse, unchanged, better, and significantly better according to the grade difference.

[0009] The present invention has the following beneficial effects: The regional comprehensive ecological environment quality evaluation method proposed in this paper constructs a new remote sensing ecological index PRSEI based on MODIS, which integrates NDVI, WET, LST, NDBWI, PM 2.5Each of the five indicators is completely based on remote sensing image acquisition and can effectively reduce noise interference. Compared with the traditional RSEI, it supplements the indicators that characterize the quality of the atmospheric environment and improves the comprehensiveness and representativeness of the remote sensing ecological index.

[0010] The present invention proposes a remote sensing image information granulation method based on the continuous value of indicators, and uses the idea of ​​spatial information granulation to calculate PRSEI. This method allows each remote sensing pixel and each indicator value to participate in the calculation, and establishes their own weights from the spatial distribution of each PRSEI indicator. It can better handle the fuzziness and uncertainty between remote sensing information particles, enhance the ability to process complex information in remote sensing images, and is a concrete embodiment of geographic stratification and scale transformation methods. It provides a new perspective for remote sensing spatial data calculation. The method proposed in the present invention adopts the entropy weight method to determine the indicator weights, which can objectively reflect the contribution of each factor of the model to the overall evaluation results. It extracts the spatial distribution characteristics at the micro scale one by one based on the stratification of remote sensing images and then aggregates them to the macro scale, which enhances the intuitiveness and interpretability of the evaluation results and can better reflect the complexity and diversity of the ecological environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 (a) Figure 2 (b) is the spatial analysis map of the ecological environment quality of the study area in 2001 and 2022; Figure 3 (a) Figure 3 (b) is an analysis diagram of the spatial changes of the new ecological quality parameter PRSEI level in the study area from 2001 to 2012 and from 2012 to 2022 according to the present invention. DETAILED DESCRIPTION

[0012] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Example 1

[0013] Mineral resources in the Beijing-Tianjin-Hebei region are mainly concentrated around the Taihang Mountains and the Yanshan Mountains. More than 1,300 types of minerals have been discovered, covering a wide range of fields from energy to metals to non-metals, and are well-known for their rich reserves of coal and iron ore. This implementation takes Baoding, Xingtai, Handan, Shijiazhuang, Zhangjiakou, Chengde, Tangshan and Qinhuangdao in the Beijing-Tianjin-Hebei region as examples, and the monitoring period is 2001-2022 for explanation.

[0014] like Figure 1As shown, this embodiment is a regional ecological environment quality evaluation method based on neighborhood granular fuzzy c-means clustering, comprising the following steps: S1. Based on the cloud platform, the scope of the study area was defined by uploading a vector file and set as the center of the map. The MODIS data source within the required study area was defined based on the JavaScript programming language and the data source was preprocessed. PM was obtained from the Chinese high-resolution and high-quality near-surface air pollutant dataset of the National Tibetan Plateau Data Center (http: / / data.tpdc.ac.cn). 2.5 Satellite inversion data.

[0015] Specifically, the MODIS data for the study area from 2001 to 2022 is uploaded to the cloud platform. The MODIS data sources include MOD09A1, MOD13A1, and MOD11A2. The temporal resolution of MOD13A1 data products is 16 days, and the other data products are 8 days. The spatial resolution of MOD11A2 is 1000 meters, and the others are 500 meters.

[0016] The methods for preprocessing MODIS data include: S1-1 defines the cloud removal function, projection coordinate system and pixel size based on JavaScript programming language; S1-2 uses the modified Normalized Difference Water Index (MNDWI) based on the JavaScript programming language to create a mask file for the target image, perform mask water area processing, and obtain a mask image.

[0017] S2, based on the preprocessed MODIS data, the ecological factors are extracted through JavaScript programming language, and then the ecological factors are standardized so that the value range is unified between 0 and 1. The ecological factors include green factor, humidity factor, heat factor, dryness factor, PM 2.5 Concentration factor; S3, extracting the greenness factor comprises the following steps: S3-1 calculates the vegetation index NDVI of the time series according to the MOD13A1 product in step S1, performs data quality control band detection on noise such as clouds and shadows, removes outliers, and obtains the time series NDVI data. Based on the time series NDVI data, average the time series NDVI data from June to October each year to obtain the NDVI mean image, mask and project the mean NDVI image, and obtain the greenness factor thematic atlas of the study area.

[0018] NDVI is closely related to vegetation biomass, leaf area index, vegetation coverage, etc. It is an effective indicator to characterize the coverage and growth status of surface vegetation. The calculation formula is: NDVI=( r nir -r red ) / ( r nir +r red ) in, r nir , r red They are the reflectances of the near-infrared band and the red band in the MODIS data, corresponding to the second band and the first band of the surface reflectance data, respectively.

[0019] Extracting the humidity factor includes the following steps: S3-2, according to MOD09A1 in step S1, through the data product tassel cap transformation, the product quality control band is detected for cloud shadow and other noise, and the abnormal values ​​are removed to obtain the time series WET data, and the time series WET data from June to October each year are averaged and masked and projected to obtain the humidity factor thematic map of the study area; WET can better reflect the water status and humidity of water bodies, vegetation and soil. Its calculation formula is: WET=0.1147 r red + 0.2489 r nir1 +0.2408 r blue + 0.3132 r green -0.3122 r nir2 - 0.6416 r swir1 -0.5087 r swir2 In the formula, r red , r nir1 , r blue , r green , r nir2 , r swir1 , r swir2They respectively represent the reflectance of the red band, near infrared band, blue band, green band, NIR2, SWIR1, and SWIR2 bands in the MODIS data.

[0020] Extracting the dryness factor includes the following steps: S3-3, calculate the time series building-bare soil index NDBSI according to MOD09A1 in step S1, detect noise such as clouds and shadows according to the product quality control band, remove abnormal values, obtain time series NDBSI data, average the time series NDBSI data from June to October each year, and obtain the NDBSI mean image and perform masking and projection conversion to obtain the dryness factor thematic map of the study area; NDBSI is composed of the bulk index (IBI) and the bare soil index (BI), which together affect the surface dryness. The calculation formula is: NDBSI=(IBI+SI) / 2; SI=[( r swir1 + r red )-( r nir + r blue )] / [( r swir1 + r red )+( r nir + r blue )]; ,in, r red , r nir , r blue , r green , r swir1 They represent the reflectance of red, near-infrared, blue, green, and short-wave infrared in MODIS data, respectively, corresponding to the 1st, 2nd, 3rd, 4th, and 6th bands of the surface reflectance.

[0021] Extracting the heat factor includes the following steps S3-4, calculate the land surface temperature LST of the time series according to the MOD11A2 data in step S1, remove invalid observations and outliers, take the average value of the LST data from June to October each year and perform masking, perform projection conversion and resampling, and obtain the heat factor thematic map of the study area; Surface temperature is an important indicator that affects vegetation growth and environmental changes. Its calculation formula is: LST=0.02×DN-273.15 Among them, DN represents the gray value of the MOD11A2 land surface temperature data. MOD11A2 land surface temperature data has its own quality identification band. Using the quality identification to remove low-quality invalid observations can effectively reduce the impact of noise.

[0022] S3-5. Based on the above-mentioned ecological factors, standardization is performed to eliminate the dimension and make the value range uniform within (0,1). The standardized calculation formula is: x 1=( x i - x min ) / ( x max - x min ),in, x 1 is the ecological factor value after normalization; x i is the original pixel value of the indicator; x min is the minimum value of a certain ecological factor in all years; x max It is the maximum value of a certain ecological factor in all years.

[0023] S4, based on the entropy weight method of neighborhood granulation based on the standardized ecological factors, the weighted sum of the various indicators of PRSEI is performed to obtain the remote sensing ecological index PRSEI based on MODIS; Specifically, PRSEI is an improved new remote sensing ecological index based on MODIS, which integrates the characteristics of ecosystem greenness, humidity, heat, dryness, PM 2.5 NDVI data, WET data, LST data, NDBSI data, satellite PM 2.5 Based on the five indicators of concentration inversion data, the remote sensing image pixels of various ecological factors are divided into four categories based on neighborhood granular fuzzy c-means clustering.

[0024] For the standardized data of each ecological factor obtained in step S3-5, the remote sensing image pixels of each ecological factor are divided into 4 categories based on the neighborhood granular fuzzy c-means clustering. The specific calculation steps are as follows: S4-1. Granulate the samples to form particle vectors, cluster the samples based on the distance between the particle vectors, and obtain the central membership of each sample point to all classes by optimizing the objective function to determine the class of the sample point and automatically divide the sample data.

[0025] S4-2, assign the membership function of each cluster to each sample, quantify the cluster to which the sample belongs based on the size of the membership, and further optimize the objective function to obtain the minimum value.

[0026] Specifically, the optimization objective function is updated by u ij and c j , the formula is: ,in , the distance metric selected in this experiment is the Euclidean distance metric, that is, , m is the fuzzy coefficient (FC), 1 <m< , n is the number of samples, C is the number of cluster centers, x i represents the i-th sample, c j represents the jth cluster center, u ij Represents sample x i For cluster center c j Membership u ij satisfy .

[0027] S5, based on the four categories divided by the neighborhood granular fuzzy c-means clustering algorithm, the MODIS-based remote sensing ecological index PRSEI was obtained according to the entropy weight method; Specifically, based on the images of remote sensing pixels classified for each ecological factor, the knowledge granularity, knowledge granularity entropy and scale conversion knowledge granularity entropy of each ecological factor are calculated. The weights of each ecological factor are calculated according to the scale conversion knowledge granularity entropy, and the weighted summation of the weights is used to obtain PRSEI. The specific calculation steps are as follows: S5-1. Knowledge granularity can refine each indicator into different hierarchical partitions based on different remote sensing information particles, and divide it into a spatial image composed of discrete values, thereby simplifying remote sensing spatial calculations. The larger the knowledge granularity, the more average the indicator value. The formula is as follows: ,in is the knowledge granularity, is the number of remote sensing information particles of different ecological factors, U is the number of ecological factor pixels of all remote sensing images, n is the total number of factor remote sensing information particles.

[0028] S5-2. Based on the spatial variability of each ecological factor, the knowledge granularity entropy is calculated according to the remote sensing information particles of different ecological factors. The formula is as follows: ,in , is the knowledge granularity entropy of ecological factors, i For different remote sensing information particles, j For the j Regions, is the number of each remote sensing information particle of different indicators.

[0029] S5-3. Based on the study area, different geographical regions have different natural environmental factors and different environmental policies. According to the influence of different urban overall evaluation factors, the weights of various ecological factors are calculated according to the scale conversion knowledge granularity entropy. The smaller the scale conversion knowledge granularity entropy, the finer the evaluation area division, the stronger the resolution, and the greater the contribution to PRSEI. The formula is as follows: ,in, is the knowledge granularity entropy of scale transformation, is the knowledge granularity entropy of some cities in the study area, and n is the number of regions, that is, n=8.

[0030] S5-4. The PRSEI of the study area is obtained by multiplying each ecological factor with its weight and accumulating them. The closer the PRSEI value is to 1, the better the ecological environment quality is, and the closer it is to 0, the worse the ecological environment quality is. The weight formula of each indicator is defined by the scale transformation knowledge granularity entropy as follows: ,in, i Represents different ecological factors, represents the weight of each factor, It represents the knowledge granularity entropy of scale transformation of each factor in different regions.

[0031] S6. Based on the PRSEI of different years, the ecological environment quality of the study area was evaluated according to the graded assignment method, and the difference method was used to obtain the changes in ecological environment quality between years.

[0032] Specifically, the remote sensing ecological index PRSEI in the study time series is divided into five levels using the natural breakpoint method, which are expressed as poor (0.2-0.46), poor (0.46-0.53), medium (0.53-0.59), good (0.59-0.63), and good (0.63-0.7), and reassigned to 1-5. For the changes in ecological environment quality in the two years of 2001-2012 and 2012-2022, the PRSEI grade values ​​of the most recent year and the starting year are subtracted, and the ecological quality changes are divided into five situations according to the grade difference: significantly worse, worse, unchanged, better, and significantly better. Figure 2 (a) and 2(b) are the spatial distribution maps of PRSEI in the embodiment area of ​​the present invention in 2001 and 2022, among which the PRSEI of Handan, western Xingtai, and northern Zhangjiakou is poor, while the PRSEI of Shijiazhuang, Baoding, Tangshan, and Qinhuangdao is better. Figure 3(a) and 3(b) are the spatial change diagrams of PRSEI levels in the regions of the embodiments of the present invention from 2001 to 2012 and from 2012 to 2022. The implementation results show that during the period from 2001 to 2012, the ecological environment quality in Handan, western Xingtai, central Zhangjiakou, and southeastern Chengde improved most significantly, while some areas in northern Chengde experienced a decline in PRSEI levels. From 2012 to 2022, Tangshan, Qinhuangdao, and some areas of Shijiazhuang were mostly concentrated in the transition from excellent to relatively excellent levels, and were generally still at a good ecological quality level.

[0033] Although the present invention is described herein with reference to a number of illustrative embodiments of the present invention, it will be appreciated that those skilled in the art may devise many other modifications and implementations that fall within the scope and spirit of the principles disclosed herein. More specifically, within the scope of the present disclosure, drawings, and claims, a variety of variations and modifications may be made to the components and / or layout of the subject combination layout. In addition to the variations and modifications made to the components and / or layout, other uses will also be apparent to those skilled in the art.

Claims

1. A method for evaluating regional ecological environment quality based on neighborhood granular fuzzy c-means clustering, characterized in that: The following steps are involved: Step 1: Calculate and reconstruct the remote sensing time series MODIS data to extract ecological factors; Step 2: Standardize the ecological factors; Step 3: Divide the standardized remote sensing images of each ecological factor into n categories based on neighborhood granular fuzzy c-means clustering; Step 4: The n ecological factors divided by the neighborhood granular fuzzy c-means clustering algorithm are used to obtain the MODIS-based comprehensive remote sensing ecological index PRSEI using the improved entropy weight method. The sub-steps are as follows: 4-1: Divide each ecological factor into different levels according to the knowledge granularity; 4-2: Calculate the knowledge granularity entropy based on the remote sensing information particles of different ecological factors; 4-3: Calculate the scale transformation knowledge granularity entropy based on the influence of the overall evaluation factors in different regions; 4-4: Calculate the weights of each ecological factor based on the scale transformation knowledge granularity entropy; 4-5: The PRSEI of the study area is obtained by multiplying the remote sensing images of each ecological factor with its weight and accumulating them; Step 5: Obtain the ecological environment quality of the study area and the changes in the ecological environment quality of the mining area over the years.

2. According to claim 1, a method for evaluating regional ecological environment quality based on neighborhood granular fuzzy c-means clustering is characterized in that: The MODIS data products in step 1 include MOD13A1, MOD11A2 and MOD09A1.

3. According to claim 1, a method for evaluating regional ecological environment quality based on neighborhood granular fuzzy c-means clustering is characterized in that: The MODIS data standardization process in step 2 includes defining a cloud removal function and a water body mask function for MODIS series satellite remote sensing image data.

4. The regional ecological environment quality evaluation method based on neighborhood granular fuzzy c-means clustering according to claim 1 is characterized in that: In step 3, the clustering categories are divided by the neighborhood granular fuzzy c-means clustering algorithm, which includes the following steps: Step 1: Granulate the samples to form particle vectors, cluster the samples based on the distance between the particle vectors, and obtain the central membership of each sample point to all classes by optimizing the objective function to determine the class of the sample point and automatically divide the sample data; Step 2: Assign the membership function of each cluster to each sample, quantify the cluster to which the sample belongs based on the size of the membership, and further optimize the objective function to obtain the minimum value.

5. The regional ecological environment quality evaluation method based on neighborhood granular fuzzy c-means clustering according to claim 1 is characterized in that: In step 5, based on the PRSEI of different years, the ecological environment quality of the study area is obtained according to the graded assignment method, where the graded assignment means that the PRSEI of a specific year is divided into five levels: poor, relatively poor, medium, good, and excellent by the natural breakpoint method, and the grades are re-assigned from 1 to 5; The change in ecological environment quality of the mining area within the year interval is obtained by the difference method. The difference method refers to subtracting the PRSEI grade values ​​of the starting year, and dividing the ecological quality change into five levels according to the grade difference: obvious deterioration, deterioration, unchanged, improvement, and obvious improvement.

Images