Method and system for testing water extraction accuracy based on SAR (Synthetic Aperture Radar) image

Abstract

The invention relates to a water body extraction accuracy test method and system based on SAR images, and belongs to the technical field of water body monitoring and remote sensing processing. SAR image data are automatically obtained through intelligent retrieval and preprocessed, an improved SDWI index algorithm is adopted for water body extraction, and binarization classification is achieved through an adaptive threshold method; on the basis of pixel-level statistical analysis and time sequence regression analysis, obtaining an image pixel mean value and an inter-annual change slope over the years; and finally, through image element value remapping of the interannual change slope beta and high-resolution image comparison verification, misjudgment area identification is realized. According to the method, the accuracy of water body extraction is improved, full-process automation from data acquisition, processing and analysis to result verification is also realized, and the efficiency and reliability of water body extraction precision verification are remarkably improved.

Description

Technical Field

[0001] The invention relates to a method and system for testing the accuracy of water body extraction based on SAR images, and belongs to the technical field of water body monitoring and remote sensing processing. Background Art

[0002] With the rapid development of remote sensing (RS) and geographic information systems (GIS), water extraction using satellite remote sensing imagery has become a crucial tool for water resource management and environmental monitoring. The Sentinel-1 satellite, a radar remote sensing satellite launched by the European Space Agency (ESA), provides synthetic aperture radar (SAR) data that is unrestricted by sunlight and cloud cover, offering unique advantages for water monitoring. The SDWI (Water Extraction Using Sentinel-1 SAR Data) method, a highly efficient water extraction technique, accurately identifies and extracts water areas based on Sentinel-1 SAR data, providing valuable data resources for water resource management.

[0003] However, water extraction often results in uncertainties due to differences in remote sensing imagery conditions, variations in land cover, and extraction algorithms. To ensure the accuracy and reliability of water extraction results, accurate verification and calibration are crucial. Traditional verification methods often rely on on-site field surveys or comparisons with other data sources. These methods are not only time-consuming and labor-intensive, but also struggle to fully cover the entire extraction area. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a water body extraction accuracy verification method based on SAR images, which not only improves the accuracy of water body extraction, but also realizes the automation of the entire process from data acquisition, processing and analysis to result verification, significantly improving the efficiency and reliability of water body extraction accuracy verification.

[0005] The technical solution adopted by the present invention is: The water body extraction accuracy test method based on SAR imagery includes the following steps: S1. Acquire remote sensing images and preprocess them; S2. Use the improved SDWI index method combined with adaptive threshold to extract water bodies from the pre-processed remote sensing image; S3. Perform pixel statistics on the water extraction results of multiple images over the years, arrange the water extraction results by image time, and then calculate the mean value of each grid pixel in one year, and calculate the mean value M of the image pixels over the years. i , where i is the number of years; S4. Perform pixel-level time series regression analysis to quantify the long-term trend of water body changes by calculating the interannual variation slope β of each pixel; The calculation formula is as follows: , β: interannual variation slope of each pixel; i: the i-th year; n: total number of years; M i : the mean value of image pixels in year i; S5. Remap the pixel values based on the obtained pixel trend to positive and negative values. β greater than 0 indicates a high probability of water at this grid location, while β less than 0 indicates a low probability of water at this grid location. If β = 0, the water extraction result is accurate. If β > 0 but there is no water in a certain area, or β < 0 but there is water in a certain area, the cause of the misjudgment should be investigated. The accuracy of the water extraction result can be further verified by combining optical imagery or field data.

[0006] In the above method, step S1 first automatically obtains SAR image data that meets the requirements from the Copernicus Open Access Hub platform through an intelligent retrieval tool. The retrieval content includes imaging time, resolution, coverage, data format, and projection mode. After the data is downloaded, a standardized preprocessing process including radiometric calibration, terrain correction, geometric correction, and image cropping is immediately performed. The result is a preprocessed remote sensing image in raster data format.

[0007] In step S2, the SDWI index is calculated using the data of the VH and VV polarization bands. According to the value of the SDWI index, an appropriate threshold is set to segment water bodies from non-water bodies. When the calculated SDWI result is greater than 0, it is judged to be a water body, and the rest are non-water bodies. When the object to be extracted is the water body of a reservoir, the calculation formula of the SDWI of the improved SDWI index method is as follows: , b1: band value of VV polarization, b2: Band value of VH polarization.

[0008] An adaptive threshold method is used for binary classification, and the classification results under multiple thresholds are automatically tested through an accuracy evaluation algorithm. Finally, the threshold scheme with the best classification accuracy is selected, that is, the threshold is 0.4, to complete the accurate extraction of water body elements and output the raster results. The traditional SDWI algorithm only relies on simple linear combinations (such as SDWI = VV - VH) and is easily affected by noise and surface scattering. The present invention uses an improved SDWI algorithm and adopts a collaborative analysis cross-polarization entropy calculation method to enhance the separability of water bodies and background objects. Traditional SDWI algorithms mostly use fixed thresholds, which are difficult to adapt to changes in scattering characteristics in different regions and seasons. The present invention sets the threshold to 0.4, which can more accurately distinguish water bodies from other objects.

[0009] In step S3, when focusing on the accuracy test of the water extraction results of the reservoir, the images of two adjacent months of the flood season each year are selected for calculation, M i =(V1+V2) / 2, where V1 is the pixel value of the image of the first month, and V2 is the pixel value of the image of the second month.

[0010] Another object of the present invention is to provide a water body extraction accuracy verification system based on SAR images, comprising An automatic remote sensing data acquisition module, which automatically acquires qualified SAR image data from the Copernicus Open Access Hub platform through intelligent retrieval and performs standardized pre-processing procedures including radiometric calibration, terrain correction, geometric correction, and image cropping; Water body extraction module, which calculates the water body index of each pixel through the built-in improved SDWI index algorithm engine and uses the adaptive threshold method to complete the binary classification; Pixel statistics and time series regression analysis module, which uses built-in functions to automatically perform pixel mean statistics operations and pixel-level time series regression analysis operations on water body extraction results over the years to obtain the interannual variation slope of each pixel; The extraction result accuracy test module is based on the time series regression analysis results. This module automatically remaps the pixel value of the water body change trend coefficient β (β>0 indicates a high probability of water body existence, and β<0 indicates a low probability of water body existence) to generate an intuitive classification result map. The classification result map generated after pixel value remapping (β>0 or β<0) is used to determine which area is misjudged. By comparing the high-resolution remote sensing images or field survey data of the same period, the misjudgment area can be effectively identified.

[0011] The beneficial effects of the present invention are: (1) The present invention combines the improved SDWI algorithm with the adaptive threshold optimization method to significantly improve the extraction accuracy. By introducing pixel-level time series regression analysis, the accurate quantification of water body extraction results is achieved. The accuracy of the extraction results is judged based on the quantification results. An intelligent detection mechanism based on β-value pixel remapping is constructed. The misjudgment areas are detected and corrected through the pixel remapping results, which greatly improves the accuracy of water body extraction.

[0012] (2) The present invention can carry out precise verification in specific areas, effectively reducing the time and labor costs required for traditional large-scale verification, and greatly improving the accuracy and reliability of water extraction results.

[0013] (3) The present invention realizes the automation of the entire process from data acquisition, processing and analysis to result verification through the collaborative process of "intelligent retrieval and standardized preprocessing → improved SDWI algorithm combined with adaptive threshold optimization → pixel-level time series regression analysis → pixel value remapping", significantly improving the efficiency and reliability of water body extraction accuracy verification, and providing scientific technical support for applications such as dynamic monitoring of water resources and flood disaster assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is the water body extraction result of a reservoir in the embodiment (image time August 2019); Figure 3 is the average pixel value of a certain reservoir over the years in the embodiment; Figure 4 The pixel-level time series regression analysis result of a reservoir in the embodiment; Figure 5 This is the pixel value remapping result of a reservoir in the embodiment. DETAILED DESCRIPTION

[0015] The present invention is further described below with reference to specific embodiments.

[0016] Example 1 Water body extraction accuracy test method based on SAR image, (such as Figure 1 ) includes the following steps: S1. Remote Sensing Image Acquisition and Preprocessing: First, an intelligent search tool automatically retrieves qualified SAR image data from the Copernicus Open Access Hub platform. The search criteria include imaging time, resolution, coverage, data format, and projection. Once the data is downloaded, a standardized preprocessing process is immediately performed, including radiometric calibration, terrain correction, geometric correction, and image cropping. The resulting preprocessed remote sensing image is a raster data format.

[0017] S2. Use the improved SDWI index method combined with the adaptive threshold to extract water bodies from the pre-processed remote sensing image: The preprocessed remote sensing imagery is then used to extract water bodies using the improved SDWI method (extracting water bodies based on SAR data from the Sentinel-1 satellite). This method uses data from the VH and VV polarization bands to calculate the SDWI index. Based on the SDWI index value, an appropriate threshold is set to segment water bodies from non-water bodies, thereby extracting the target water body. When the calculated SDWI result is greater than 0, it is considered a water body, and the rest are non-water bodies. After repeated experiments, the SDWI calculation formula for extracting reservoir water bodies is as follows: Formula (1); b1: Band value of VV polarization b2: Band value of VH polarization.

[0018] This embodiment automatically obtains image data of a reservoir from July to August 2019 to 2023 based on retrieval, performs preprocessing operations on it, calculates the improved SDWI index through built-in functions, and uses the adaptive threshold method to complete the binary classification and extract the water body results, such as Figure 2 As shown, the black area is the water body.

[0019] S3. Perform pixel statistics on the water extraction results of multiple images over the years, arrange the water extraction results by image time, and then calculate the mean value of each grid pixel in one year, and calculate the mean value M of the image pixels over the years. i , where i is the number of years: Based on the water body results extracted using the SDWI method, pixel statistics of multiple images over the years are performed, where the statistical value is the average of the image pixels over the years. Since this invention focuses on the accuracy test of the water body extraction results of the reservoir, images of two adjacent months during the flood season each year are selected for calculation. i =(V1+V2) / 2, where V1 is the pixel value of the image of the first month, and V2 is the pixel value of the image of the second month.

[0020] The pixel mean value of each year for a certain reservoir is calculated, and 5 sets of values are obtained from 2019 to 2023, such as Figure 3 shown.

[0021] S4. Perform pixel-level time series regression analysis to quantify the long-term trend of water body changes by calculating the interannual variation slope β of each pixel; In order to quantify the long-term trend of water body changes, pixel-level time series regression analysis can be performed to determine the spatial distribution characteristics of the possibility of water body existence by calculating the interannual variation slope (β) of each pixel. The calculation formula of β is based on the basic linear trend (least squares method), which is as follows: Formula (2); β: interannual variation slope of each pixel; i: the i-th year; n: total number of years; M i : Mean value of image pixels in year i.

[0022] A pixel-level time series regression analysis was conducted on a certain reservoir to obtain the interannual variation slope β value of each pixel, such as Figure 4 shown.

[0023] S5. Remap the pixel values based on the obtained pixel trend to positive and negative values. β greater than 0 indicates a high probability of water at this grid location, while β less than 0 indicates a low probability of water at this grid location. If β = 0, the water extraction result is accurate. If β > 0 but there is no water in a certain area, or β < 0 but there is water in a certain area, the cause of the misjudgment should be investigated. The accuracy of the water extraction result can be further verified by combining optical imagery or field data.

[0024] Based on the time series regression analysis results of a reservoir, the accuracy of the extraction results was tested and the pixel value of the water body change trend coefficient β was remapped. The results are as follows Figure 5 β>0 indicates a high probability of water body existence, and β<0 indicates a low probability of water body existence. Figure 5 In the black area in the lower half, the water body extraction result is water body, but β<0 in this place, indicating that the probability of water body existence is very low, which indicates that there may be misjudgment. By comparing the high-resolution remote sensing images or field survey data of the same period, the misjudgment area can be effectively identified.

[0025] Example 2: A water body extraction accuracy verification system based on SAR images, comprising: An automatic remote sensing data acquisition module, which automatically acquires qualified SAR image data from the Copernicus Open Access Hub platform through intelligent retrieval and performs standardized pre-processing procedures including radiometric calibration, terrain correction, geometric correction, and image cropping; Water body extraction module, which calculates the water body index of each pixel through the built-in improved SDWI index algorithm engine and uses the adaptive threshold method to complete the binary classification; after receiving the data preprocessing completion instruction, the system automatically starts the improved SDWI index algorithm engine, calculates the water body index of the image pixel by pixel, and then uses the adaptive threshold method to perform binary classification. The classification results under multiple thresholds are automatically tested through the system's built-in accuracy evaluation algorithm, and finally the threshold scheme with the best classification accuracy is selected, that is, the threshold is 0.4, to complete the accurate extraction of water body elements and output the raster results.

[0026] The pixel statistics and time series regression analysis module uses built-in functions to automatically perform pixel mean statistical operations and pixel-level time series regression analysis operations on the water body extraction results to obtain the inter-annual variation slope of each pixel. After receiving the water body extraction processing completion instruction, the system automatically starts the built-in function to perform pixel statistical operations on multiple images over the years, and the output result is the image pixel mean data M over the years. i , where i is the number of years; the output data format is raster data; after receiving the mean value data of image pixels over the years, the system automatically starts the built-in function (2) to perform time series regression analysis, and the output result is the inter-annual variation slope β of each pixel; The extraction result accuracy test module automatically remaps the pixel value of the water body change trend coefficient β based on the time series regression analysis results (β>0 indicates a high probability of water body existence, β<0 indicates a low probability of water body existence), generates an intuitive classification result map, and judges which place is misjudged based on the classification result map generated after pixel value remapping (β>0 or β<0 map); after receiving the interannual change slope of each pixel, the system automatically remaps the pixel value of the water body change trend coefficient β (β>0 indicates a high probability of water body existence, β<0 indicates a low probability of water body existence), generates an intuitive classification result map, and the output data format is raster data.

[0027] The system adopts a modular design, integrating functions such as data processing, algorithm calculation, visual analysis and report generation. It supports end-to-end automated processing from raw data input to final output, significantly improving the efficiency and accuracy of dynamic monitoring of reservoir water bodies.

[0028] The above is a further description of the present invention in conjunction with specific embodiments, and the protection scope of the present invention is not limited thereto.

Claims

1. A water body extraction accuracy test method based on SAR images is characterized by: The steps are as follows: S1. Acquire remote sensing images and preprocess them; S2. Use the improved SDWI index method combined with adaptive threshold to extract water bodies from the pre-processed remote sensing image; S3. Perform pixel statistics on the water extraction results of multiple images over the years, arrange the water extraction results by image time, and then calculate the mean value of each grid pixel in one year, and calculate the mean value M of the image pixels over the years. i , where i is the number of years; S4. Perform pixel-level time series regression analysis to quantify the long-term trend of water body changes by calculating the interannual variation slope β of each pixel. The calculation formula is as follows: ， β: interannual variation slope of each pixel, i: the i-th year, n: total number of years, M i : the mean value of image pixels in year i; S5. Remap the pixel values based on the obtained pixel trend according to positive and negative values. If β = 0 in the area, the water extraction result is accurate. If β > 0 in a certain area but there is actually no water, or β < 0 in a certain area but there is actually water, the cause of the misjudgment should be investigated. Combined with optical imagery or field data verification, the accuracy of the water extraction result can be further determined.

2. The water body extraction accuracy testing method based on SAR imagery according to claim 1 is characterized in that: In step S1, the intelligent retrieval tool is used to automatically retrieve qualified SAR image data from the Copernicus Open Access Hub platform. The retrieval content includes imaging time, resolution, coverage, data format, and projection mode. After the data is downloaded, a standardized preprocessing process including radiometric calibration, terrain correction, geometric correction, and image cropping is immediately performed.

3. The water body extraction accuracy testing method based on SAR imagery according to claim 1 is characterized in that: In the improved SDWI index method described in step S2, when the extraction object is a water body of a reservoir, the calculation formula of SDWI is as follows: ， b1: band value of VV polarization, b2: Band value of VH polarization.

4. The water body extraction accuracy verification method based on SAR imagery according to claim 1, wherein: In step S3, when focusing on the accuracy test of the water extraction results of the reservoir, the images of two adjacent months of the flood season each year are selected for calculation, M i =(V1+V2) / 2, where V1 is the pixel value of the image of the first month, and V2 is the pixel value of the image of the second month.

5. The water extraction accuracy verification system based on SAR images is characterized by: include An automatic remote sensing data acquisition module, which automatically acquires qualified SAR image data from the Copernicus Open Access Hub platform through intelligent retrieval and performs standardized pre-processing procedures including radiometric calibration, terrain correction, geometric correction, and image cropping; Water body extraction module, which calculates the water body index of each pixel through the built-in improved SDWI index algorithm engine and uses the adaptive threshold method to complete the binary classification; Pixel statistics and time series regression analysis module, which uses built-in functions to automatically perform pixel mean statistics operations and pixel-level time series regression analysis operations on water body extraction results over the years to obtain the interannual variation slope of each pixel; The extraction result accuracy test module is based on the results of time series regression analysis. This module automatically remaps the pixel values of the water body change trend coefficient β to generate an intuitive classification result map. If β>0 in a certain area but there is actually no water, or β<0 but there is actually water, it is necessary to check the cause of the misjudgment. By comparing the high-resolution remote sensing images or field survey data of the same period, the misjudgment area can be effectively identified.

Images