A water body extraction method and device based on improved maximum inter-class variance method

Through the improved maximum inter-class variance method combined with multiple treatment methods, the problem of degradation of water extraction accuracy and noise interference caused by radar shadow in SAR remote sensing images is solved, and a higher water extraction accuracy is achieved.

CN118521917BActive Publication Date: 2024-11-08SHANDONG FENGSHI INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410985043.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-08
Estimated Expiration
2044-07-23

Smart Images

  • Figure CN118521917B_ABST
    Figure CN118521917B_ABST
Patent Text Reader

Abstract

The present invention relates to a water body extraction method and equipment based on an improved maximum inter-class variance method, and belongs to the technical field of remote sensing data processing. Using the VV and VH polarization SAR image data of the ascending and descending orbits as the basis, after fusion, the single polarization water body extraction method, the dual polarization water body index method, and the principal component analysis method are used to extract the water body, and the extraction results are superimposed and selected for expansion processing, and iterative processing is performed through principal component analysis and inter-class variance calculation. The method of the present invention solves the problem of poor water body extraction accuracy caused by the confusion between radar shadows and water bodies, and effectively reduces the interference of SAR image noise on water body extraction, thereby improving the water body extraction accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a water body extraction method, in particular to a water body extraction method and equipment based on an improved maximum inter-class variance method, belonging to the technical field of remote sensing data processing. Background Art

[0002] Water is the source of life and an important resource for the existence and development of life. Timely grasping the spatiotemporal changes of water resources is of great significance to human production and life. Remote sensing images have become an important data source in the field of water body information extraction due to their advantages of fast information acquisition speed, short cycle and wide range. The water body extraction method based on SAR remote sensing images is based on the fact that the scattering rate of water bodies to microwaves is lower than that of other land objects, which makes the water bodies form a clear contrast with the surrounding land or vegetation in SASR images. At the same time, SAR sensors can penetrate clouds and obtain high-resolution surface information under different weather conditions. Compared with water body extraction based on optical remote sensing images, it has certain advantages. However, due to the certain tilt of the satellite sensor during the shooting process, the SAR image data will have some radar shadow areas in the obtained images, making the backscattering coefficient of the shadow closer to the backscattering coefficient of the water body, which confuses the extraction results and leads to a decrease in extraction accuracy.

[0003] Therefore, the idea of ​​combining ascending and descending SAR remote sensing image data for water body extraction came into being. Fusion of ascending and descending SAR images can solve the confusion problem of water body extraction caused by radar shadow to a certain extent. However, the fusion of multi-view images brings more speckle noise. Using only single polarization data for water body extraction cannot effectively eliminate the influence of noise and is easily interfered by vegetation and buildings. The traditional dual-polarization water body index method cannot effectively solve the noise interference. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a water body extraction method based on an improved maximum inter-class variance method, taking ascending and descending orbit SAR remote sensing image data as the basic data for water body extraction, and adopting a combination of single polarization water body extraction method, water body index method, principal component analysis method and maximum inter-class variance method and other processing methods to perform water body extraction, solve the problem of poor water body extraction accuracy caused by confusion between radar shadow and water body, and effectively reduce the interference of SAR image noise on water body extraction, thereby improving water body extraction accuracy.

[0005] The technical solution adopted by the present invention is:

[0006] A water body extraction method based on an improved maximum inter-class variance method comprises the following steps:

[0007] S1. Acquire SAR remote sensing image data and preprocess it;

[0008] S2. Perform polarization processing on the preprocessed data to obtain VV polarization SAR image data and VH polarization SAR image data of the ascending and descending orbits, and fuse the two ascending and descending orbit SAR images;

[0009] S3. The data after the fusion of ascending and descending SAR images are used to extract water bodies using the single polarization threshold method, dual polarization water index and principal component analysis method respectively: The single polarization threshold method water body extraction uses the VV polarization backscattering coefficient and VH polarization backscattering coefficient after the fusion of ascending and descending SAR images. AD VV , AD VH , respectively draw a "double-peak" histogram of the backscatter coefficient, extract the backscatter coefficient value at the bottom of the valley between the "double peaks" as the segmentation threshold of water and non-water objects, and extract water bodies; dual-polarization water index water body extraction is to extract water bodies by calculating the dual-polarization water index after the fusion of ascending and descending SAR images; principal component analysis water body extraction is to perform principal component transformation on the data after the fusion of ascending and descending SAR images, select the principal component with the largest amount of information, draw a "double-peak" histogram of the backscatter coefficient, and determine the segmentation threshold by the backscatter coefficient value at the bottom of the valley to extract water bodies;

[0010] S4. The water pixels extracted by the three methods in step S3 are assigned a value of 1, and the parts without water are uniformly assigned a value of 0, and the extraction results of the three methods are superimposed on the same grid position, and the part of the superimposed result greater than or equal to X is selected, and the expansion extraction process is performed;

[0011] S5. Perform principal component transformation on the data extracted in step S4 to select principal component data with large information content, calculate the inter-class variance of water bodies and non-water bodies, select the data with the largest inter-class variance as the extraction result, and then expand the extraction again based on the extraction result. Iterate through principal component analysis and inter-class variance calculation, and the result of each iterative extraction is clipped with the clipping range. After multiple iterations, the result with the smallest absolute value of the area difference between two iterations is selected as the final result.

[0012] In the above method,

[0013] The fusion process of the ascending and descending SAR images in step S2 is as follows: the two images of the ascending and descending orbits with VV polarization are added, and the two images of the ascending and descending orbits with VH polarization are added.

[0014] The VV polarization backscatter coefficient and the VH polarization backscatter coefficient after the fusion of the ascending and descending SAR images in step S3 are respectively AD VV , AD VH The calculation formula is:

[0015] ,

[0016] ,

[0017] In the formula is the backscatter coefficient of the VV polarization ascending SAR image, is the backscatter coefficient of the descending SAR image with VV polarization, is the backscatter coefficient of the VH polarization ascending SAR image, is the backscatter coefficient of the down-orbit SAR image with VH polarization.

[0018] In step S3, the dual-polarization water index sets the segmentation threshold of water body extraction to near 0. When the dual-polarization water index is greater than 0, it is a water body, and when it is less than 0, it is a non-water body. The specific calculation formula of the dual-polarization water index is as follows:

[0019] .

[0020] In step S4, in the single polarization threshold method water body extraction, the water body in the two extraction results is assigned to 1 and the non-water body is assigned to 0; X=3 is selected; the expansion extraction processing is to use a small expansion factor to morphologically expand it to obtain a range for subsequent clipping, and use a large expansion factor to perform morphological expansion to obtain a buffer zone.

[0021] The inter-class variance calculation formula of water bodies and non-water bodies in step S5 is as follows:

[0022] ,

[0023] Where G is the between-class variance between water bodies and non-water bodies, is the proportion of water pixels in the image, is the proportion of non-water pixels in the image, is the average gray value of the water body, is the average gray value of non-water bodies, is the average gray value of the entire image.

[0024] Another object of the present invention is to provide a water body extraction device based on an improved maximum inter-class variance method, comprising a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the water body extraction method based on the improved maximum inter-class variance method as described above.

[0025] The beneficial effects of the present invention are:

[0026] (1) Combining ascending and descending orbit SAR image data for water extraction can effectively solve the problem of low backscatter coefficient of ground objects caused by radar shadow, which affects the water extraction results;

[0027] (2) Using the multi-water body extraction method combined with the maximum inter-class variance method to extract water bodies can solve the noise problem caused by multi-scene image fusion and improve the accuracy of water body extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of the method of the present invention;

[0029] Figure 2 Schematic diagram of principal component analysis for water extraction;

[0030] Figure 3 The principal component analysis-maximum inter-class variance iteration flow chart of the present invention. DETAILED DESCRIPTION

[0031] The present invention is further described below in conjunction with specific embodiments.

[0032] Embodiment 1 A water body extraction method based on the improved maximum inter-class variance method comprises the following steps (see Figure 1 )as follows:

[0033] S1. Obtain SAR remote sensing image data and preprocess it:

[0034] The sample data mainly includes the Level-1 GRD products of the Sentinel-1 A satellite in IW mode, and these SAR data are processed by orbit correction, thermal noise removal, speckle filtering, multi-viewing, geometric correction and radiation calibration to eliminate the radiation distortion and geometric distortion caused by disturbances caused by a series of factors such as the atmospheric environment, the earth's rotation, and random noise, correct the distortion of the geometric shape or position in the original image, and reduce the impact of errors.

[0035] S2. Perform polarization processing on the preprocessed data to obtain ascending and descending orbit VV polarization SAR image data and ascending and descending orbit VH polarization SAR image data, and fuse the two ascending and descending orbit SAR images:

[0036] In the VV and VH polarized SAR remote sensing images, water bodies have smaller backscattering coefficients than other land objects, which is manifested in the image as the relatively "darker" color of the water bodies. This feature makes the histogram of the backscattering coefficient of the SAR remote sensing image appear to be "double-peaked", and the backscattering coefficient value at the bottom of the valley between the "double peaks" is the segmentation threshold for water bodies and non-water objects.

[0037] The fusion process of ascending and descending SAR images adopts the band calculation pixel superposition method. The two images of ascending and descending orbits with VV polarization are added, and the two images of ascending and descending orbits with VH polarization are added.

[0038] For the radar shadow problem in the image, the following formula can be used to calculate the VV and VH polarization backscatter coefficients after the ascending and descending SAR image fusion: AD VV , AD VH , and then extract the water body segmentation threshold.

[0039] ,

[0040] ,

[0041] In the formula is the backscatter coefficient of the VV polarization ascending SAR image, is the backscatter coefficient of the descending SAR image with VV polarization, is the backscatter coefficient of the VH polarization ascending SAR image, is the backscatter coefficient of the down-orbit SAR image with VH polarization.

[0042] S3. Use the data after the fusion of ascending and descending SAR images to perform water extraction using single polarization threshold method, dual polarization water index method and principal component analysis method respectively:

[0043] The single polarization threshold method for water extraction is based on the VV and VH polarization backscattering coefficients after the fusion of ascending and descending SAR images. AD VV , AD VH , draw a "double-peak" histogram of the backscattering coefficient, extract the backscattering coefficient value at the bottom of the valley between the "double peaks", and use it as the segmentation threshold between water bodies and non-water objects to extract water bodies.

[0044] Water body extraction with dual polarization water index can use VV and VH polarization data of Sentinel-1 to extract water body at the same time, or the segmentation threshold of water body extraction can be attributed to near 0. When the dual polarization water index is greater than 0, it is water body, and when it is less than 0, it is non-water body. The specific calculation formula is as follows:

[0045] ,

[0046] When the value of 0 cannot distinguish water bodies from non-water bodies well in some special cases, a “double peak” graph can be drawn to obtain the segmentation threshold for water body extraction.

[0047] Sentinel-1 IW mode GRD data contains two polarizations: VV and VH. Although the information content of dual-polarization data is higher than that of single-polarization data, which is more conducive to image classification and interpretation, there is a certain correlation between the two polarization modes, and there is data redundancy. The principal component analysis method can retain the main information through function transformation, reduce the amount of data, and achieve the purpose of enhancing or extracting useful information.

[0048] Here, the ascending and descending orbits can be imaged in dual polarization. X As a two-dimensional space, X i Represents the pixel value of the pixel. A is the transformation matrix, that is X The transpose of the matrix of eigenvectors of the covariance matrix.

[0049] ,

[0050] ,

[0051] ,

[0052] After principal component transformation Y Contains two principal components, the first principal component y 1 concentrates the largest amount of information, such as Figure 2 , and the second principal component y 2 The amount of information is very small, mainly noise. y The information of 1 can better distinguish water bodies and non-water bodies. y The calculation formula for part 1 is as follows:

[0053] ,

[0054] After obtaining the calculated image, draw its "double peak" graph, and determine the segmentation threshold through the backscattering coefficient value of the valley bottom to extract the water body.

[0055] S4. The water pixels extracted by the three methods in step S3 are assigned a value of 1 respectively, and the parts without water are uniformly assigned a value of 0. The extraction results of the three methods are superimposed on the same grid position, and the parts of the superimposed results greater than or equal to X are selected, and the expansion extraction process is performed:

[0056] In the process of comparing the extraction results of various methods, the present invention finds that the distribution of noise points in different water body extraction results is different. Therefore, the method first uses ascending and descending orbit single polarization water body extraction, dual polarization water body extraction, principal component analysis and other methods to perform preliminary extraction of water body information on the preprocessed satellite remote sensing image, and the extracted water body pixels are respectively assigned 1, and the parts without water body are uniformly assigned 0. Then, the extraction results at the same grid position are superimposed, and the part of the superimposed result greater than or equal to X is selected. Here, X=3 is selected. In this way, the result obtained may be smaller than the actual water body area. Therefore, it is necessary to use a smaller expansion factor to perform morphological expansion to obtain a range for subsequent clipping. At the same time, a larger expansion factor is used to perform morphological expansion to obtain a buffer zone.

[0057] S5. The data extracted in step S4 is transformed into principal components to select principal component data with large information content, and the inter-class variance of water bodies and non-water bodies is calculated. The largest inter-class variance is selected as the extraction result, and then the extraction is expanded again based on the extraction result. The principal component analysis and inter-class variance calculation are iteratively processed, and the result of each iterative extraction is clipped with the clipping range. After multiple iterations, the result with the smallest absolute value of the area difference between two iterations is selected as the final result:

[0058] The maximum inter-class variance method divides the SAR image into two parts according to a certain threshold, which are defined as foreground and background. The effect of the division of the two parts on the image is measured by the inter-class variance between them. The larger the value, the better the division. Conversely, the greater the probability of misclassification or omission. The inter-class variance calculation formula for background and foreground is as follows:

[0059] ,

[0060] Where G is the inter-class variance between foreground and background, is the proportion of foreground pixels in the image, is the ratio of background pixels in the image, is the average gray value of the foreground, is the average gray value of the background, is the average gray value of the entire image, where the foreground is the water body and the background is the non-water body.

[0061] The data extracted in step S4 is transformed into principal components to select principal component data with large information content, and the inter-class variance of water bodies and non-water bodies is calculated. The largest inter-class variance is selected as the extraction result, and then the extraction is expanded again based on the extraction result. After the extraction result is obtained through principal component analysis and inter-class variance calculation, it is iterated and processed again. After that, the extraction result is expanded again based on the extraction result, and the result of each iterative extraction is clipped with the clipping range. After multiple iterations, the one with the smallest absolute value of the area difference between two iterations is selected as the final result. The iterative processing flow is as follows: Figure 3 shown.

[0062] Embodiment 2 A water body extraction device based on an improved maximum inter-class variance method comprises a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the water body extraction method based on the improved maximum inter-class variance method as described in the above embodiment 1 is implemented.

[0063] The above description is only a specific implementation of the present application, and the protection scope of the present invention is not limited thereto.

Claims

1. A water body extraction method based on an improved maximum inter-class variance method, characterized in that: The steps include: S1. Acquire SAR remote sensing image data and preprocess it; S2. Perform polarization processing on the preprocessed data to obtain VV polarization SAR image data and VH polarization SAR image data of the ascending and descending orbits, and fuse the two ascending and descending orbit SAR images; S3. The data after the fusion of ascending and descending SAR images are used to extract water bodies using single polarization threshold method, dual polarization water index method and principal component analysis method respectively: Single polarization threshold method water body extraction is to use the VV polarization backscattering coefficient and VH polarization backscattering coefficient after the fusion of ascending and descending SAR images. AD VV , AD VH , draw a double-peaked histogram of the backscatter coefficient, extract the backscatter coefficient value at the bottom of the valley between the double peaks, and use it as the segmentation threshold of water and non-water objects to extract water bodies; dual-polarization water index water body extraction is to extract water bodies by calculating the dual-polarization water index after the fusion of ascending and descending SAR images. When the dual-polarization water index is greater than 0, it is a water body, and when it is less than 0, it is a non-water body. The specific calculation formula of the dual-polarization water index is as follows: ; The principal component analysis method for water extraction is to perform principal component transformation on the data after the fusion of ascending and descending SAR images, select the principal component with the largest amount of information, draw a bimodal histogram of the backscatter coefficient, and determine the segmentation threshold through the backscatter coefficient value at the bottom of the valley to extract water. S4. The water pixels extracted by the three methods in step S3 are assigned a value of 1 respectively, and the parts without water are uniformly assigned a value of 0. The extraction results of the three methods are superimposed on the same grid position, and the part of the superimposed result greater than or equal to X is selected, and the expansion extraction process is performed, X=3; S5. Perform principal component transformation on the data extracted in step S4 to select principal component data with large information content, calculate the inter-class variance of water bodies and non-water bodies, select the data with the largest inter-class variance as the extraction result, and then expand the extraction again based on the extraction result. Iterate through principal component analysis and inter-class variance calculation, and the result of each iterative extraction is clipped with the clipping range. After multiple iterations, the result with the smallest absolute value of the area difference between two iterations is selected as the final result.

2. The water body extraction method based on the improved maximum inter-class variance method according to claim 1 is characterized in that: The fusion process of the ascending and descending SAR images in step S2 is as follows: the two images of the ascending and descending orbits with VV polarization are added, and the two images of the ascending and descending orbits with VH polarization are added.

3. The water body extraction method based on the improved maximum inter-class variance method according to claim 1 is characterized in that: The VV polarization backscatter coefficient and the VH polarization backscatter coefficient after the fusion of the ascending and descending SAR images in step S3 are respectively AD VV , AD VH The calculation formula is: , , In the formula is the backscatter coefficient of the VV polarization ascending SAR image, is the backscatter coefficient of the descending SAR image with VV polarization, is the backscatter coefficient of the VH polarization ascending SAR image, is the backscatter coefficient of the down-orbit SAR image with VH polarization.

4. The water body extraction method based on the improved maximum inter-class variance method according to claim 1 is characterized in that: The expansion extraction process described in step S4 is to use a small expansion factor to perform morphological expansion to obtain a range for subsequent clipping, and use a large expansion factor to perform morphological expansion to obtain a buffer zone.

5. The water body extraction method based on the improved maximum inter-class variance method according to claim 1 is characterized in that: The inter-class variance calculation formula of water bodies and non-water bodies in step S5 is as follows: , Where G is the between-class variance between water bodies and non-water bodies, is the proportion of water pixels in the image, is the proportion of non-water pixels in the image, is the average gray value of the water body, is the average gray value of non-water bodies, is the average gray value of the entire image.

6. A water extraction device based on an improved maximum inter-class variance method, comprising a processor and a memory storing computer program instructions; wherein: When the processor executes the computer program instructions, the water body extraction method based on the improved maximum between-class variance method as described in any one of claims 1 to 5 is implemented.