A Geometric Correction Method for High-Orbit Staring SAR Based on Multi-Angle Weighted Images of Mountainous Areas

Abstract

This invention provides a geometric correction method for high-orbit staring SAR based on multi-angle weighted image simulation in mountainous areas. The method involves selecting a DEM image D for the corresponding region; determining the coordinates of the four corner points of image D; calculating the corresponding simulated image value for each pixel in image D; performing multi-scale decomposition on each SAR multi-angle image using a non-downsampling pyramid to obtain high-frequency and low-frequency sub-bands for each image; fusing the high-frequency and low-frequency sub-bands using Gaussian blur; calculating the fusion coefficients of the high-frequency and low-frequency sub-bands; and then determining the grayscale value of the fused image; finally, matching the simulated image with the SAR image using the SIFT algorithm to obtain the geometrically corrected image. This invention effectively alleviates the problem of the significant impact of elevation on the geometric correction accuracy of images during high-orbit SAR imaging in mountainous areas.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, specifically a high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas. Background Technology

[0002] High-orbit (GEO) synthetic aperture radar (SAR) staring mode is an imaging mode that utilizes the significant curvature of the GEO SAR orbit to perform multiple imaging operations from multiple angles on a specific area, thereby acquiring more information about the current scene. Existing technologies perform image simulation and geometric correction processing on single-view SAR images. The information obtained from a fixed single viewpoint is limited, and it cannot extract target scattering features that can only be observed from other viewpoints. GEO staring SAR imaging, with its high orbital altitude and significant orbital curvature, allows for multi-angle imaging of the same scene. However, currently there are no methods for processing and extracting information from multi-angle staring SAR images for geometric correction. Furthermore, due to the complex terrain of mountainous areas, imaging exhibits more overlapping and shadowing phenomena compared to plains areas, significantly impacting the accuracy of image localization.

[0003] Geometric correction is a crucial step in SAR image processing and an important process for applying SAR imagery to global and regional information systems. The primary objective of SAR image data processing is to obtain geodetic reference spatial coordinate information through the establishment of a geostationary model. High-orbit SAR images possess advantages such as large swath width and multiple revisits, offering significant advantages for geographic scene detection. Furthermore, high-orbit staring SAR, by performing multiple imaging operations on the same scene from multiple azimuth angles, can further extract geographic information. Therefore, through precise geometric correction of high-orbit staring SAR images, we can obtain geodetic reference spatial coordinate information, enabling comprehensive analysis of multi-source and multi-temporal information. Currently, existing technologies for processing multi-view images mainly include the following schemes:

[0004] Image simulation and geometric correction of single-view SAR images based on backscattering models is currently the most widely used method. It mainly involves solving the RD positioning model of the image to obtain image information such as slant range, then using an indirect positioning method combined with the backscattering model to perform image simulation, and finally performing geometric fine correction of mountain images through image registration. Although this method is widely used in geometric correction processing of low-orbit strip mode images, the impact of elevation changes on imaging is greater in high-orbit SAR images when the orbital altitude increases and the orbital curvature becomes larger, resulting in a decrease in the accuracy of geometric correction using traditional methods.

[0005] Existing technologies primarily obtain image information such as slant range by solving the RD positioning model of the image, and then use indirect positioning methods combined with backscattering models for image simulation. In the image simulation process, a global DEM elevation model with an accuracy of 30m is typically used. However, due to further improvements in imaging accuracy, the accuracy of the global DEM model no longer meets the requirements of current imaging accuracy. Therefore, it is necessary to further correct the elevation model using staring SAR multi-angle images to reduce the dependence on traditional elevation models during geometric correction, thereby improving the accuracy of correction. Existing technologies are mainly widely used in geometric correction processing of single-view mode images. Single-view SAR systems suffer from severe information loss and poor image interpretation. High-orbit staring SAR utilizes the curved orbit of a high-orbit system to conduct multi-azimuth observations. By repeatedly observing the same area from different azimuth angles, multi-view geometric and scattering information of ground features within that area can be obtained. Combining this with the method adopted in this invention can effectively improve the accuracy of geometric correction. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas. Specifically, it uses a backscattering model and DEM data to simulate and generate SAR images of the scene from multi-angle staring SAR images. Gaussian weighted information fusion is performed on multiple simulated images, and finally, geometric correction of the high-orbit SAR image is achieved through a matching algorithm. This method overcomes the shortcomings of existing technologies and can be widely applied.

[0007] The purpose of this invention is to provide a high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas, comprising the following steps:

[0008] Step 1: Acquire multi-angle high-orbit staring SAR images, and select the corresponding DEM image based on the image localization results. .

[0009] Step 2: Obtain the image using the indirect positioning method. The coordinates of the four corner points , , and

[0010] Step 3: For the image Each pixel Use the following steps:

[0011] (3a) Determine the rectangular projection coordinates of the pixel and its four surrounding pixels. After coordinate transformation, the ECF coordinate vector of the current target pixel is obtained. The ECF coordinate vectors of the four pixels are named clockwise from top to bottom as follows: , , and .

[0012] (3b) Calculate the normal vector of the plane based on the target pixel vector. , , , The calculation formula is as follows (using...). (For example)

[0013] , , ,

[0014] in, , , ,

[0015] (3c) Calculate the average value of the four normal vectors.

[0016] (3d) For the current pixel The RD model is used for positioning calculation to obtain the SAR simulated image coordinates of the corresponding pixels.

[0017]

[0018] in, The GEI coordinate vector to be determined for the current pixel. The equatorial radius of the Earth's ellipsoid. The polar radius of the Earth's ellipsoid. For Doppler frequency, This represents the elevation value of the current cell in the DEM. For radar signal wavelength, This is the satellite's current position vector. This is the satellite's current velocity vector. , .

[0019] (3e) The GEI coordinate vector of the current cell in (3d) and satellite current position vector The vector from the target to the radar is calculated. ;

[0020] (3f) Calculate the local incident angle of the corresponding pixel. and radar cross section :

[0021]

[0022]

[0023] (3h) Radar cross section from (3f) Calculate the current DEM pixels The corresponding simulated image values.

[0024] Step 4: Perform multi-scale decomposition on each SAR multi-angle image using a non-downsampled pyramid. The structure of each layer of the non-downsampled pyramid is shown below, where... To input multi-angle images, To output the decomposed image, the high-frequency subband is obtained. and low-frequency subband ,in It is the number of decomposition layers. It is the image number. , , , These are dual-channel non-subsampling filters, and satisfy... .

[0025]

[0026] Step 5: A shearing filter bank is used to decompose the SAR multi-angle image of each layer into images in various directions. The structure of the filter bank is shown in the figure below. To localize the output direction of the image, , , , They are shear filters, and satisfy... .

[0027]

[0028] Step 6: Calculate the mean value of the low-frequency subband coefficients and the mean of high-frequency subband coefficients ,according to Calculate the coefficient of each low-frequency subband. arrive Euclidean distance and variance ,according to Calculate the coefficient of each high-frequency subband arrive Euclidean distance and variance The calculation formula is as follows:

[0029] , ,

[0030] ,

[0031] ,

[0032] in, The total number of images. This represents the number of layers in the image decomposition.

[0033] Step 7: Calculate the fused low-frequency sub-band according to the Gaussian blur calculation rules. and the fused high-frequency subband The calculation formula is as follows;

[0034]

[0035]

[0036] in, For the first Weights of the low-frequency subband of a SAR image. For the first Weights of high-frequency subbands in a SAR image;

[0037] Step 8: Add the fused low-frequency subband to the fused high-frequency subband to obtain the fused image. ;

[0038] Step 9: Match the simulated image with the SAR image using the SIFT algorithm to obtain the geometrically corrected image. .

[0039] This invention effectively compensates for the lack of information and low geometric correction accuracy of single-view SAR images by fusing multi-angle image information of staring SAR, and effectively alleviates the problem that elevation has a significant impact on the geometric correction accuracy of images when imaging in mountainous areas with high-orbit SAR. Attached Figure Description

[0040] Figure 1 This is a flowchart of the present invention; Detailed Implementation

[0041] The flowchart of this invention is as follows: Figure 1 As shown.

[0042] A geometric correction method for high-orbit staring SAR based on multi-angle weighted image simulation in mountainous areas includes the following steps:

[0043] Step 1: Acquire multi-angle high-orbit staring SAR images, and select the corresponding DEM image based on the image localization results. .

[0044] Step 2: Obtain the image using the indirect positioning method. The coordinates of the four corner points , , and

[0045] Step 3: For the image Each pixel Use the following steps:

[0046] (3a) Determine the rectangular projection coordinates of the pixel and its four surrounding pixels. After coordinate transformation, the ECF coordinate vector of the current target pixel is obtained. The ECF coordinate vectors of the four pixels are named clockwise from top to bottom as follows: , , and .

[0047] (3b) Calculate the normal vector of the plane based on the target pixel vector. , , , The calculation formula is as follows (using...). (For example)

[0048] , , ,

[0049] in, , , ,

[0050] (3c) Calculate the average value of the four normal vectors.

[0051] (3d) For the current pixel The RD model is used for positioning calculation to obtain the SAR simulated image coordinates of the corresponding pixels.

[0052]

[0053] in, The GEI coordinate vector to be determined for the current pixel. The equatorial radius of the Earth's ellipsoid. The polar radius of the Earth's ellipsoid. For Doppler frequency, This represents the elevation value of the current cell in the DEM. For radar signal wavelength, This is the satellite's current position vector. This is the satellite's current velocity vector. , .

[0054] (3e) The GEI coordinate vector of the current cell in (3d) and satellite current position vector The vector from the target to the radar is calculated. ;

[0055] (3f) Calculate the local incident angle of the corresponding pixel. and radar cross section :

[0056]

[0057]

[0058] (3h) Radar cross section from (3f) Calculate the current DEM pixels The corresponding simulated image values.

[0059] Step 4: Perform multi-scale decomposition on each SAR multi-angle image using a non-downsampled pyramid. The structure of each layer of the non-downsampled pyramid is shown below, where... To input multi-angle images, To output the decomposed image, the high-frequency subband is obtained. and low-frequency subband ,in It is the number of decomposition layers. It is the image number. , , , These are dual-channel non-subsampling filters, and satisfy... .

[0060]

[0061] Step 5: A shearing filter bank is used to decompose the SAR multi-angle image of each layer into images in various directions. The structure of the filter bank is shown in the figure below. To localize the output direction of the image, , , , They are shear filters, and satisfy... .

[0062]

[0063] Step 6: Calculate the mean value of the low-frequency subband coefficients and the mean of high-frequency subband coefficients ,according to Calculate the coefficient of each low-frequency subband. arrive Euclidean distance and variance ,according to Calculate the coefficient of each high-frequency subband arrive Euclidean distance and variance The calculation formula is as follows:

[0064] , ,

[0065] ,

[0066] ,

[0067] in, The total number of images. This represents the number of layers in the image decomposition.

[0068] Step 7: Calculate the fused low-frequency sub-band according to the Gaussian blur calculation rules. and the fused high-frequency subband The calculation formula is as follows;

[0069]

[0070]

[0071] in, For the first Weights of the low-frequency subband of a SAR image. For the first Weights of high-frequency subbands in a SAR image;

[0072] Step 8: Add the fused low-frequency subband to the fused high-frequency subband to obtain the fused image. ;

[0073] Step 9: Match the simulated image with the SAR image using the SIFT algorithm to obtain the geometrically corrected image. .

[0074] The method used in this invention utilizes the multi-angle illumination characteristic of staring mode to solve the problem of large errors caused by inaccurate elevation models when generating high-orbit SAR simulation images.

[0075] The multi-angle image weighting method used in this invention to simulate images significantly reduces the computational load compared to traditional information fusion methods.

Claims

1. A geometric correction method for high-orbit staring SAR based on multi-angle weighted image simulation in mountainous areas, characterized in that, Includes the following steps: Step 1: Acquire multi-angle high-orbit staring SAR images, and select the corresponding DEM image based on the image localization results. ; Step 2: Obtain the image using the indirect positioning method. The coordinates of the four corner points , , and ; Step 3: For the image Each pixel Calculate the corresponding simulated image values: Step 4: Perform multi-scale decomposition on each SAR multi-angle image using a non-downsampled pyramid to obtain high-frequency subbands. and low-frequency subband ,in It is the number of decomposition levels. It is the image number; Step 5: Use a shearing filter bank to perform multi-directional decomposition on the SAR multi-angle image of each layer to obtain images in each direction. ; Step 6: Calculate the mean value of the low-frequency subband coefficients and the mean of high-frequency subband coefficients ,according to Calculate the coefficient of each low-frequency subband. arrive Euclidean distance and variance ,according to Calculate the coefficient of each high-frequency subband arrive Euclidean distance and variance ; Step 7: Calculate the fused low-frequency sub-band according to the Gaussian blur calculation rules. and the fused high-frequency subband ; In step seven, the fused high-frequency subband and the fused low-frequency subband The calculation formula is as follows: ， ， in, For the first Weights of the low-frequency subband of a SAR image. For the first Weights of high-frequency subbands in a SAR image; Step 8: Add the fused low-frequency subband to the fused high-frequency subband to obtain the fused image. ; Step 9: Match the simulated image with the SAR image using the SIFT algorithm to obtain the geometrically corrected image. .

2. The high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas according to claim 1, characterized in that, Step 3 includes the following sub-steps: (3a) Determine the rectangular projection coordinates of the pixel and its four surrounding pixels. After coordinate transformation, the ECF coordinate vector of the current target pixel is obtained. The ECF coordinate vectors of the four pixels are named clockwise from top to bottom as follows: , , and ; (3b) Calculate the normal vector of the plane based on the target pixel vector. , , , The calculation formula is as follows: ， ， ， ， in, , , , ; (3c) Calculate the average value of the four normal vectors. ; (3d) For the current pixel The RD model is used for positioning calculation to obtain the SAR simulated image coordinates of the corresponding pixels. , ， in, The GEI coordinate vector to be determined for the current pixel. The equatorial radius of the Earth's ellipsoid. The polar radius of the Earth's ellipsoid. For Doppler frequency, This represents the elevation value of the current cell in the DEM. For radar signal wavelength, This is the satellite's current position vector. This is the satellite's current velocity vector. , ; (3e) The GEI coordinate vector of the current cell in (3d) and satellite current position vector The vector from the target to the radar is calculated. ; (3f) Calculate the local incident angle of the corresponding pixel. and radar cross section : ， ， (3h) Radar cross section from (3f) Calculate the current DEM pixels The corresponding simulated image values.

3. The high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas according to claim 1, characterized in that, The structure of the non-downsampled pyramid in step four is composed of , , , It consists of four dual-channel non-subsampled filters, and satisfies .

4. The high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas according to claim 1, characterized in that, The structure of the shear filter bank in step five is as follows: , , , It consists of four shear filters, and satisfies .

5. The high-orbit staring SAR geometric correction method based on multi-angle weighted image simulation in mountainous areas according to claim 1, characterized in that, In step six, the calculation formula is as follows: ， ， ， ， in, The total number of images. This represents the number of layers in the image decomposition.

Images