A method for selecting flat areas on the lunar surface based on fractal dimension and polarization decomposition
By combining fractal dimensions and polarization decomposition technology, radar data is used to identify flat areas of the moon's surface, solving the problem of insufficient information acquisition and resolution of existing methods in polar regions, and achieving high-accuracy flat areas of the moon's surface.
Patent Information
- Application Number
- CN202210114176.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-30
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-01-30
AI Technical Summary
Existing methods have problems in finding flat areas of the moon that optical images cannot provide polar region information and the low resolution of digital elevation models, making it difficult to accurately identify flat regions on the moon's surface, especially in permanent shadow areas in polar regions.
Using a method based on fractal dimensions and polarization decomposition technology, the flat area of the moon's surface is screened by calculating the fractal dimensions of the total power image of the radar backscatter and decomposing the Yamaguchi four elements.
Without relying on additional DEM data, the accuracy of flat area selection is significantly improved, and the geological characteristics of the moon's surface can be effectively identified, providing support for future application research on polarized SAR data.
Smart Images

Figure CN114509758B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of radar image processing, in particular to a method for selecting a lunar surface flat area based on fractal dimension and polarization decomposition. Background Art
[0002] Synthetic Aperture Radar (SAR) is an active microwave imaging system with all-day, all-weather and penetrating imaging capabilities. It is widely used in earth remote sensing and planetary exploration. Polarimetric synthetic aperture radar provides polarization scattering information in addition to the amplitude and phase characteristics of the target scattered echo, providing an effective way to infer the physical properties of the planetary surface. Polarimetric target decomposition is an important branch of polarimetric SAR image interpretation, which decomposes the matrix that characterizes the polarization scattering characteristics of the target and extracts various polarization parameters with physical significance. Yamaguchi four-element decomposition is an incoherent polarization decomposition method, which is widely used in the identification and classification of scattering mechanisms of ground targets.
[0003] Fractal theory is a powerful mathematical tool for describing irregular geometric forms, among which fractal dimension (D) is an important parameter for measuring surface roughness based on fractal methods. For two-dimensional surfaces, its value is between 2.0 and 3.0. The fractal dimension of a smooth two-dimensional surface is 2.0. As the surface roughness increases, the fractal dimension value increases. For very rough surfaces, the fractal dimension value can reach 3.0. Therefore, the fractal dimension can be used to describe different land cover characteristics, and there is a one-to-one correspondence between the fractal dimension and the texture characteristics (or roughness) of the geological surface.
[0004] The selection of flat areas is an important requirement in lunar exploration. For example, in the exploration of water ice in the permanent shadow areas of the lunar polar regions, it is necessary to find relatively flat areas to eliminate the interference of surface roughness and provide support for the landing and sampling of future permanent shadow area landers. In addition, the flat areas on the lunar surface can also be used as distributed targets for the calibration of future lunar polarimetric SAR data. The existing methods for finding flat areas on the moon are: obtaining lunar surface roughness information through inversion of high-resolution optical images (lunar orbit optical camera or lunar rover camera); obtaining the digital elevation model (DEM) of the lunar surface through inversion of lunar orbit laser altimeter data, and then obtaining flat areas. However, optical images cannot provide information about the permanent shadow areas of the lunar polar regions, and the coverage of lunar rover images is small, covering only a very small number of areas of interest on the lunar surface. In addition, due to the inherent laser spot spacing, the existing lunar surface digital elevation model (DEM) data has low resolution in some areas of interest. Summary of the invention
[0005] In order to extract the flat areas of the area of interest on the lunar surface, the present invention proposes a method for selecting flat areas in polarimetric SAR images that combines fractal dimension (spatial geometric features) and polarimetric decomposition technology (polarimetric scattering features). The fractal dimension and polarimetric decomposition technology can be used together to select flat areas by identifying the texture features and polarimetric scattering features of the lunar surface. The method proposed in the present invention can select flat areas on natural geological surfaces using polarimetric SAR data without relying on additional DEM data, providing support for future research on the application of full polarimetric SAR data on the moon.
[0006] The method can be divided into two main steps: rough selection and fine selection. The first is the rough selection stage: the fractal dimension is calculated using the radar backscatter total power image, and then the area where the fractal dimension is less than the fractal dimension mean is extracted. Fine selection: The polarimetric SAR image is decomposed by Yamaguchi four elements, and the area with the largest proportion of single scattering power is further screened out in the candidate area roughly selected. Specifically, the technical solution of the present invention is: a method for selecting flat areas on the lunar surface based on fractal dimension and polarization decomposition, comprising the following steps:
[0007] Step 1: Acquire full polarization SAR data;
[0008] Step 2, calculating the fractal dimension D according to the total backscatter power of the radar, and obtaining a fractal dimension graph;
[0009] Step 3, binarize the fractal dimension map, set the pixel values greater than the mean value of the fractal dimension to white, and set the pixel values less than the mean to black, to obtain a binarized fractal dimension map;
[0010] Step 4: For the binarized fractal dimension image, all continuous N×N pixel regions that do not contain white pixels are screened out as roughly selected candidate regions, where N is the width of the pixel region.
[0011] Step 5, using the Yamaguchi four-element decomposition method to perform polarization decomposition on the SAR echo to obtain the magnitudes of single scattering component, secondary scattering component, volume scattering component and spiral scattering component;
[0012] Step 6: For each candidate area, select the n×n pixel area with the largest single scattering component, which is the selected flat area, where n < N.
[0013] Beneficial effects:
[0014] The present invention aims at the full polarization echo data acquired by the lunar orbit SAR, firstly calculates the fractal dimension of the backscattered total power image, and then selects the continuous N×N pixel areas less than the mean value of the fractal dimension as candidate areas; then uses the Yamaguchi four-element decomposition method to polarize the radar echo; then for each candidate area, selects the continuous n×n pixel area with the largest single scattering component as the final selected flat area. The present invention has three advantages:
[0015] (1) For areas such as the permanent shadows at the lunar poles where optical sensors cannot provide observation information, a method for selecting flat areas using polarimetric SAR data is provided;
[0016] (2) Comprehensively utilizing the texture characteristics and polarization scattering characteristics of the geological surface significantly improves the accuracy of flat area selection;
[0017] (3) Provide support for the application research of polarimetric SAR data on the moon in the future. In addition, the proposed method was verified using the full polarimetric SAR data of the Gaofen-3 satellite in the desert area and the SRTM1 DEM data. The results show that without relying on additional DEM data, the proposed method can accurately extract the geological features of the selected area and effectively realize the identification of flat areas. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 is a flow chart of the method of the present invention;
[0019] Figure 2 This is a schematic diagram of surface area calculation using the triangular prism method;
[0020] Figure 3 Schematic diagram of different grid step sizes of the local triangular prism method; (a) step size is 1; (b) step size is 2; (c) step size is 4; (d) step size is 8;
[0021] Figure 4 (a) Backscattered total power diagram;
[0022] Figure 4 (b) Fractal dimension diagram, pixels with fractal dimension D greater than its mean are set to white, and pixels with fractal dimension D less than the mean are set to black;
[0023] Figure 5 The N×N candidate regions that meet the search criteria (marked by white boxes); (a) The white solid line box marks the 200×200 pixel region searched in the fractal dimension map; (b) The magnified fractal dimension map; (c) The magnified backscattered total power map;
[0024] Figure 6The final screened flat area of 40×40 pixels (320m×320m) (white dotted box) is shown in Figure 1, and the white solid rectangular box is the roughly selected area screened according to the fractal dimension.
[0025] Figure 7 It is a scatter density map of the elevation standard deviation and the maximum elevation deviation of all 40×40 pixel windows in the entire SAR image scene. DETAILED DESCRIPTION
[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of the present invention.
[0027] To achieve the above object, the technical solution of the present invention is as follows Figure 1 As shown in FIG. 1 , a method for selecting lunar surface flat areas based on fractal dimension and polarization decomposition includes the following steps:
[0028] Step 1: Acquire full polarization SAR data;
[0029] Step 2, calculating the fractal dimension D according to the total backscatter power of the radar, and obtaining a fractal dimension graph;
[0030] Step 3, binarize the fractal dimension map, set the pixel values greater than the mean value of the fractal dimension to white, and set the pixel values less than the mean to black, to obtain a binarized fractal dimension map;
[0031] Step 4: For the binarized fractal dimension image, all continuous N×N pixel regions that do not contain white pixels are screened out as roughly selected candidate regions, where N is the width of the pixel region;
[0032] Step 5, using the Yamaguchi four-element decomposition method to perform polarization decomposition on the SAR echo to obtain the magnitudes of single scattering component, secondary scattering component, volume scattering component and spiral scattering component;
[0033] Step 6: For each candidate area, select the n×n pixel area with the largest single scattering component, which is the selected flat area, where n < N.
[0034] According to an embodiment of the present invention, the step 2 calculates the fractal dimension based on the total radar backscatter power to obtain a fractal dimension graph; specifically as follows: first, the local triangular prism method is used to calculate the fractal dimension of the radar backscatter total power image. This method regards the grayscale value of each pixel on the image as a "height" value, and then uses a 9×9 window to move on the image, and calculates the surface area under different grid steps at the center of each window. The surface area is calculated by introducing a central pixel point. The central pixel point divides the grid square into 4 triangles, and the pixel value of the central pixel point is the average of the 4 corner pixels. The sum of the areas of the 4 triangles is calculated separately to be the surface area of the grid, such as Figure 2 As shown. A, B, C, D are grid pixels, A', B', C', D' are the corresponding grayscale values, P is the center point of the square, and the pixel value of point P is the average of the pixel values of the four corner points.
[0035] By changing the grid step size (the step sizes are 1, 2, 4, and 8 respectively), repeating the above calculations, we can obtain the corresponding relationship between the grid surface area and the step size, as shown in Figure 3 The following is a schematic diagram of different grid step sizes of the local triangular prism method. From left to right, the step sizes of (a), (b), (c), and (d) are 1, 2, 4, and 8, respectively. In the logarithmic coordinate system (with base 2), the surface area and step size are fitted by least squares to obtain the slope of the fitted line. k , and then use the following formula to calculate the fractal dimension D at the center of the window:
[0036] D =2- k (1)
[0037] Then move the window to the next pixel and repeat the above process until the entire image is calculated.
[0038] According to one embodiment of the present invention, the fractal dimension is calculated by taking the backscattered total power image of the fully polarized SAR data of GF-3 in the desert area as an example. The size of the selected SAR data is 6197×8035 pixels, and the data number is: GF3_KAS_QPSI_007178_E83.8_N39.9_20171221_L1A_AHV_L10002865872. The resolution of the image is 8m, the central longitude is 83°47'21"E, and the latitude is 39°56'36"N.
[0039] Step 3: Binarize the fractal dimension map, set pixel values greater than the mean value of the fractal dimension to white, and set pixel values less than the mean value to black, to obtain a binarized fractal dimension map.
[0040] The calculated fractal dimension D is averaged, and pixels greater than the average are set to 255 (white), and pixels less than the average are set to 0 (black). The larger the fractal dimension, the greater the terrain undulation and the rougher the surface; the smaller the fractal dimension, the relatively flat terrain undulation and the smoother the surface.
[0041] Figure 4 (a) shows the total backscattered power map, and Figure 4 (b) shows the binarization of the calculated fractal dimension D image, where pixels greater than the average value of the fractal dimension D are set to 255 (white), and pixels less than the average value are set to 0 (black). It can be seen from Figures 4 (a) and 4 (b) that the texture features of the fractal dimension map can basically reflect the texture features of the total power map, which is brighter in hilly areas and other areas with large fluctuations, and darker in relatively flat areas. Therefore, the fractal dimension calculated by the total backscattered power can be used to "roughly select" flat areas on the surface.
[0042] Step 4: For the binarized fractal dimension image, all continuous N×N pixel regions that do not contain white pixels are screened out as roughly selected candidate regions, where N is the pixel region width; specifically, the following steps are performed:
[0043] According to the fractal dimension graph, the largest continuous area N×N that does not contain white pixels is selected as the candidate area of the flat area. The specific implementation method is to use an N×N sliding window to slide on the image, and check whether the sum of the pixel values inside the window is always less than 1 after each sliding. If it is satisfied, the position information of the sliding window at this time is recorded. If not, continue to slide down until all searches are completed.
[0044] Assuming N=200, the fractal dimension graph of Figure 4 is screened pixel by pixel, and multiple N×N pixel candidate regions that meet the conditions can be obtained, such as Figure 5 As shown, from Figure 5 It can be seen that the searched area (white solid rectangular box) is displayed as a uniform area of radar backscattering in the backscatter coefficient diagram (c), without obvious terrain undulations, which indicates that the rough selection of flat areas based on the fractal dimension D of the backscatter coefficient is reliable.
[0045] Step 5, using the Yamaguchi four-element decomposition method to perform polarization decomposition on the SAR echo to obtain the magnitudes of single scattering component, secondary scattering component, volume scattering component and spiral scattering component;
[0046] The calculation of the scattered component is realized by the Yamaguchi four-element decomposition method, which converts the polarization coherence matrix of the ground object into T It is decomposed into four scattering mechanisms: single scattering, secondary scattering, volume scattering and helical scattering components, as shown in the following formula:
[0047] (2)
[0048] in, is the Pauli scattering vector, with superscript represents the conjugate transpose, S is the target scattering coefficient, S HH , S VV , S HV Subscript H and V represent horizontal and vertical polarization respectively, f s , f d , f v and f c is the expansion coefficient to be determined. Single scattering is characterized by the Bragg surface scattering model, and its coherence matrix has the following form:
[0049] (3)
[0050] Secondary scattering is characterized by the dihedral scattering model, and its coherence matrix is:
[0051] (4)
[0052] in, α and β is the coefficient to be determined, | α |<1,| β |<1.
[0053] Coherence matrix for volume scattering model T v Will be based on Depending on the size of , we can choose a dipole model that obeys different probability density functions. The coherence matrix of the randomly oriented dipole model is as follows:
[0054] (5)
[0055] The coherence matrix for the dipole model following a sinusoidal distribution is given by:
[0056] (6)
[0057] When -2 dB < P < 2 dB, the volume scattering model is T v1 ;when P < -2 dB, the volume scattering model is T v2 ;when P > 2 dB, the volume scattering model isT v3 .
[0058] For the spiral scattering component with rotation invariance, its coherence matrix can be expressed as:
[0059] (7)
[0060] The above GF3 data are subjected to Yamaguchi four-element polarization decomposition, where j is an imaginary unit.
[0061] Step 6: For each candidate area, select the n×n pixel area with the largest single scattering component, which is the selected flat area, where n < N.
[0062] Flat areas are mainly dominated by single scattering components. Therefore, the fine selection is to search according to the proportion of each scattering component power of each pixel in the roughly selected candidate area, and find the continuous area (n×n pixels) with the largest proportion of single scattering power, which is the final selected flat area.
[0063] The pixels in the candidate area are searched in an n×n pixel area according to the proportion of each scattering component. The specific implementation method is to use an n×n sliding window to slide on the image, and calculate the sum of the single scattering power of all pixels inside the sliding window every time it slides until the entire candidate area is slid. Finally, the sliding window position with the largest proportion of single scattering power is recorded, which is the final flat area selected.
[0064] Considering that the resolution of the GF3 full polarization image is 8m×8m, we set n=40 here, and finally get the selected area as shown below (white dotted rectangle, 320m×320m), which has the largest single scattering power ratio among the candidate areas (white realization rectangle), reaching 88.7%. Figure 6 , the final screened 40×40 pixel (320m×320m) flat area (white dashed box 2). The white solid rectangular box 1 is the rough selection area screened according to the fractal dimension. Figure 7 It is a scatter density map of the elevation standard deviation and the maximum elevation deviation of all 40×40 pixel windows in the entire SAR image scene.
[0065] In order to verify the effectiveness of the proposed method, the elevation information of the area is analyzed using 30m resolution SRTM1 DEM data. First, the DEM data and SAR images are geocoded, and then the maximum elevation deviation and elevation standard deviation in all 40×40 pixel windows in the entire scene are calculated to obtain the scatter density distribution map as shown in the figure. Figure 7The maximum elevation deviation of the flat area (the area shown in the white dotted box) selected by the method proposed in the present invention is 3.39 m, and the elevation standard deviation is 0.72 m. Therefore, the method proposed in the present invention can effectively select flat areas on natural geological surfaces without relying on additional DEM data.
[0066] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, and it should be clear that the present invention is not limited to the scope of the specific embodiments, for those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.
Claims
1. A method for selecting lunar surface flat areas based on fractal dimension and polarization decomposition, characterized in that: The steps include: Step 1: Acquire full polarization SAR data; Step 2: Calculate the fractal dimension D according to the total backscatter power of the radar to obtain a fractal dimension graph, including: Firstly, the local triangular prism method is used to calculate the fractal dimension of the radar backscatter total power image, which is: Step 2.1, the gray value of each pixel on the image is regarded as the height value; Step 2.2, then use a window to move on the image, and calculate the surface area under the current grid step at the center of each window; Step 2.3, change the grid step size, and repeat step 2.2 to obtain the corresponding relationship between the grid surface area and the step size; Step 2.4, perform least square fitting on the obtained surface area and step length to obtain the slope of the fitting line; Step 2.5, then calculate the fractal dimension at the center of the window; Step 2.6, move the window to the next pixel and repeat the process of steps 2.2-2.5 until the entire image is calculated; Step 3, binarize the fractal dimension map, set the pixel values greater than the mean value of the fractal dimension to white, and set the pixel values less than the mean to black, to obtain a binarized fractal dimension map; Step 4: For the binarized fractal dimension image, all continuous N×N pixel regions that do not contain white pixels are screened out as roughly selected candidate regions, where N is the width of the pixel region. Step 5, using the Yamaguchi four-element decomposition method to perform polarization decomposition on the SAR echo to obtain the magnitudes of single scattering component, secondary scattering component, volume scattering component and spiral scattering component; Step 6: For each candidate area, select the n×n pixel area with the largest single scattering component, which is the selected flat area, where n < N, n is the pixel width of the selected pixel area, including: The pixels in the candidate area are searched in an n×n pixel area according to the proportion of each scattering component. An n×n sliding window is used to slide on the image. The sum of the single scattering powers of all pixels inside the sliding window is calculated each time the sliding window slides until the entire candidate area is slid. Finally, the sliding window position with the largest proportion of single scattering power is recorded, which is the final selected flat area.
2. The method for selecting lunar surface flat areas based on fractal dimension and polarization decomposition according to claim 1 is characterized in that: The step 3 specifically includes: The mean of the calculated fractal dimension D is calculated, and pixels greater than the mean are set to white, and pixels less than the mean are set to black.
3. The method for selecting lunar surface flat areas based on fractal dimension and polarization decomposition according to claim 1 is characterized in that: The step 4 specifically includes: According to the fractal dimension graph, the largest continuous area N×N that does not contain white pixels is selected as the candidate area for the flat area; an N×N sliding window is used to slide on the image to check whether the sum of the pixel values inside the window is always less than 1 after each sliding. If it is satisfied, the position information of the sliding window at this time is recorded; if not, it continues to slide down until the entire search is completed.
