A SAR image offset tracking method not affected by patch errors

Abstract

The application provides a SAR image offset tracking method which is not affected by patch errors, comprising the following steps: step 1, data preprocessing; step 2, obtaining a cross-correlation coefficient matrix related to a pixel to be measured in a main image based on a statistical method and an INFLO algorithm; step 3, calculating offset in a range direction and an azimuth direction by normalizing the peak position of the cross-correlation coefficient matrix and a SAR single-view complex image pixel; and step 4, performing steps 2-3 pixel by pixel, so as to obtain a ground surface offset map covering the regions of two SAR single-view complex images. The application can effectively remove patch errors and reduce the loss of normal offset, thereby effectively improving the precision of SAR image offset tracking results, and can adapt to SAR image data of different types and different regions, and has stronger practicability.

Description

TECHNICAL FIELD

[0001] The application relates to a SAR image offset tracking method not affected by patch errors and belongs to the field of ground surface deformation monitoring. BACKGROUND

[0002] Precise acquisition of ground surface displacement plays an important supporting role in monitoring and preventing disasters induced by factors such as glacier movement, landslides and mining activities. As an active microwave remote sensing technology, space-borne synthetic aperture radar (SAR) has the ability of all-weather and all-day observation, and can continuously obtain high-quality images under complex weather conditions, thereby providing a data basis for continuous ground surface displacement estimation. Ground surface displacement estimation methods based on SAR images mainly include differential interferometric SAR (D-InSAR) technology and offset tracking technology. The D-InSAR technology and its derivative technologies can realize millimeter-level precision displacement monitoring by extracting the interferometric phase change of the same position within a certain interval. However, in areas with a long time interval, unstable surface scattering information or large displacement, the interferogram is prone to lose coherence, which makes it difficult to apply the D-InSAR technology.

[0003] The offset tracking technology can obtain two-dimensional displacement information in the range and azimuth directions by matching two SAR intensity images of the same area obtained at different times. This method is not sensitive to the loss of interferometric phase coherence and can obtain large-scale displacement caused by factors such as glacier movement, landslides and earthquake faults. It is one of the important means for monitoring large-scale ground surface displacement. In the area where InSAR fails, the offset tracking technology shows good complementary ability and application prospect.

[0004] In the existing SAR image offset tracking method, the cross-correlation matching calculation is usually based on the correlation coefficient index to measure the similarity of the pixels in the window involved in the cross-correlation matching. When there is a pixel with intensity value significantly higher than the neighborhood statistical level in the cross-correlation matching window, because its intensity value is too different from the window mean value and standard deviation, it will dominate in the correlation calculation process, thereby causing deviation in the matching result. Such pixels are intensity value abnormal pixels. When the same intensity value abnormal pixel is repeatedly included in multiple adjacent windows and participates in the matching calculation, a regular patch artifact with the same shape as the cross-correlation matching window will appear in the offset tracking result, that is, patch error, which seriously affects the accuracy and reliability of the offset result. The current methods for removing patch error in SAR image offset tracking results mainly include threshold method and weighted cross-correlation method. The threshold method is based on the statistical characteristics that the SAR intensity image approximately obeys Rayleigh distribution, constructs a functional relationship between the scale parameter and the intensity threshold by minimizing the matching error, and sets a fixed threshold to identify the intensity value abnormal pixels in the SAR intensity image. This method is easy to implement and has certain patch error suppression capability, but the statistical acquisition of the scale parameter needs to rely on the distribution characteristics of the intensity image, and its adaptability to different SAR data and study areas is weak; in addition, in the process of removing patch error, this method will inevitably eliminate some pixels with normal scattering intensity, which will adversely affect the continuity and accuracy of the subsequent displacement field. The weighted cross-correlation method first performs logarithmic transformation on the SAR intensity image to make its pixel value distribution approximate to normal distribution, and then uses the absolute median deviation criterion to detect intensity abnormal points; for the detected intensity value abnormal pixels, a weighted normalized cross-correlation algorithm is used to suppress their influence. This method has good performance in patch error suppression, but its weighted cross-correlation calculation process is relatively complex and has low calculation efficiency; in addition, if there are many intensity value abnormal pixels or they are distributed in patches, directly setting their weights to 0 will greatly reduce the number of pixels participating in the matching, and the normal offset result will be lost. SUMMARY

[0005] The purpose of the present application is to provide a SAR image offset tracking method that is not affected by patch error to solve the problems existing in the prior art.

[0006] In order to solve the above technical problems, the present application provides a SAR image offset tracking method that is not affected by patch error, comprising: step 1, data preprocessing: obtaining two SLC SAR single-view complex images of the same coverage area, which are respectively set as the main image and the auxiliary image; registering the main image and the auxiliary image based on external DEM, and resampling the auxiliary image to the coordinate system of the main image.

[0007] Step 2, obtaining the cross-correlation coefficient matrix related to the to-be-detected pixel in the main image based on a statistical method and an INFLO algorithm, comprising: step 2.1, setting a target window and a matching window: setting a target window with the to-be-detected pixel as the center in the main image, and setting a matching window with the same size as the target window and with the to-be-detected pixel as the center in the auxiliary image; step 2.2, oversampling the SLC pixels in the target window and the matching window respectively, and then calculating the SAR intensity value based on the SLC pixel values after oversampling; for the target window and the matching window at this time: the two windows participating in the calculation of the cross-correlation coefficient matrix are called cross-correlation windows; (1) identifying candidate intensity value abnormal pixels in the cross-correlation window by using a statistical method; (2) identifying intensity value abnormal pixels by using the INFLO algorithm based on the candidate intensity value abnormal pixels, and replacing the intensity value of the intensity value abnormal pixel with the intensity median value in the neighborhood window of the intensity value abnormal pixel; (3) calculating the normalized cross-correlation coefficient matrix in the cross-correlation window.

[0008] Step 3, calculating the offset amount in the range direction and the azimuth direction through the peak position of the normalized cross-correlation coefficient matrix and the SAR single-view complex image pixel.

[0009] Step 4, performing steps 2-3 pixel by pixel, so as to obtain a complete ground surface offset map covering the regions of the two SAR single-view complex images.

[0010] In a specific embodiment, in step 2.2, the INFLO algorithm is used to identify the intensity value abnormal pixels based on the candidate intensity value abnormal pixels, and the intensity value of the intensity value abnormal pixel is replaced with the intensity median value in the neighborhood window of the intensity value abnormal pixel, specifically: (1) calculating the forward k-neighborhood of the candidate intensity value abnormal pixel: for each candidate intensity value abnormal pixel p, using the k-neighbor method to find its k nearest neighbors, the set of all k-nearest neighbors of the candidate intensity value abnormal pixel p is its forward k-neighborhood, denoted as N k (p); (2) calculating the reverse k-neighborhood of the candidate intensity value abnormal pixel: calculating the reverse k-neighborhood of each candidate intensity value abnormal pixel p through its forward k-neighborhood N k (p), denoted as RN k (p); (3) calculating the INFLO value of the candidate intensity value abnormal pixel p and identifying the intensity value abnormal pixel: the union set of the forward k-neighborhood and the reverse k-neighborhood of each candidate intensity value abnormal pixel p constitutes the influence space SN k (p) of each candidate intensity value abnormal pixel p, SN k (p) = N k (p)∪RN k(p), the INFLO value of each candidate intensity value abnormal pixel p is calculated by affecting space, and the candidate intensity value abnormal pixel p with the INFLO value greater than a preset INFLO threshold value is an intensity value abnormal pixel; (4) replacing the intensity value of the intensity value abnormal pixel: for each intensity value abnormal pixel, a 7*7 neighborhood window is taken with the intensity value abnormal pixel as the center, and the intensity value of the intensity value abnormal pixel is replaced by the intensity median in the neighborhood window.

[0011] In a specific embodiment, in step 2, the statistical method is a Z-score method, a standardized value Z of the intensity value corresponding to each pixel in the cross-correlation window is obtained by a Z-score algorithm, and the pixel with the standardized value Z greater than a preset threshold value is set as a candidate intensity value abnormal pixel, and the calculation formula of the standardized value Z of the intensity value of each pixel is as follows: , wherein x is the intensity value of the pixel in the cross-correlation window, μ is the mean of the intensity values of all pixels in the cross-correlation window, and σ is the standard deviation of the intensity values of all pixels in the cross-correlation window.

[0012] In a specific embodiment, the INFLO value of each candidate intensity value abnormal pixel p is calculated as follows: (1) the reachable density of each candidate intensity value abnormal pixel p is calculated: , wherein k distance(p) is the distance of the kth nearest neighbor of p; (2) the INFLO value of each candidate intensity value abnormal pixel p is calculated: , avg density (SN k (p)) is the average value of the reachable densities of all pixels in SN k (p), and INFLO(p) is the INFLO value of the candidate intensity value abnormal pixel p.

[0013] In a specific embodiment, step 3 is specifically: taking the position in the matrix where the peak value in the cross-correlation coefficient matrix is located, and the offset of the position from the center position of the cross-correlation coefficient matrix is the integral pixel offset of the to-be-detected pixel; performing correlation coefficient surface fitting according to the cross-correlation coefficient matrix, and the sub-pixel offset of the to-be-detected pixel in the main image can be obtained by solving the extreme value of the fitting function; and the offsets in the distance direction and the azimuth direction are obtained by comprehensively considering the integral pixel offset and the sub-pixel offset.

[0014] In a specific embodiment, in step 1, two SAR single-view complex images are registered, and the method is as follows: a conversion relationship lookup table between SAR coordinates and geographic coordinates is generated by using the track parameters and imaging geometry parameters of the two SAR single-view complex images and combining the external DEM in the SAR single-view complex image coverage area; the external DEM in the SAR single-view complex image coverage area is converted to the SAR coordinate system by using the lookup table; and the main and auxiliary images are registered, and the auxiliary image is resampled to the main image coordinate system so as to keep consistent with the main image in the geographic space.

[0015] In a specific embodiment, in step 2.2, the normalized cross-correlation coefficient matrix in the cross-correlation window is calculated, and specifically, the SAR intensity value is converted from the spatial domain to the frequency domain by Fourier transform, the cross-correlation coefficient of each position in the cross-correlation window is calculated, and the normalized cross-correlation coefficient matrix is obtained.

[0016] Compared with the prior art, the present application has the following beneficial effects.

[0017] 1. The present application quickly screens out candidate intensity value abnormal pixels by a statistical method, identifies intensity value abnormal pixels, normal pixels such as feature edges of ground objects based on the candidate intensity value abnormal pixels by the INFLO algorithm, considers the position information of the image pixels themselves, improves the ability to distinguish intensity value abnormal pixels and normal pixels such as feature edges of ground objects, and effectively removes patch errors and reduces the loss of normal offset, thereby effectively improving the tracking result precision of the SAR influence offset.

[0018] 2. The present application does not depend on SAR image scale parameter estimation and can adapt to SAR image data of different types and different regions, and has stronger practicability. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 It is a flowchart of the SAR image offset tracking method of the present application which is not affected by patch errors.

[0020] Figure 2 It is a flowchart of the cross-correlation calculation of the target window and the matching window.

[0021] Figure 3 It is a flowchart of identifying intensity value abnormal pixels and replacing the intensity values thereof.

[0022] Figure 4 It is a schematic diagram of the monitoring result by using a traditional method.

[0023] Figure 5 It is a schematic diagram of the monitoring result of the embodiment of the present application. DETAILED DESCRIPTION

[0024] The application will be described in detail below with reference to the embodiments and drawings. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0025] Reference Figure 1 A SAR image offset tracking method not affected by patch errors, comprising: step 1, data preprocessing: obtaining two SLC SAR single-view complex images of the same coverage area, which are set as a main image and an auxiliary image; registering the main image and the auxiliary image based on external DEM, and resampling the auxiliary image to the main image coordinate system.

[0026] Preferably, in step 1, the main and auxiliary SAR single-view complex images are registered by the following method: using the orbit parameters and imaging geometric parameters of the two SAR single-view complex images, and combining the external DEM in the coverage area of the SAR single-view complex images to generate a conversion relationship lookup table between the SAR coordinates and the geographic coordinates; converting the external DEM in the coverage area of the SAR single-view complex images to the SAR coordinate system through the lookup table; combining the external DEM converted to the SAR coordinate system to register the main and auxiliary images, and resampling the auxiliary image to the main image coordinate system to make it consistent with the main image in the geographic space.

[0027] Optionally, to ensure the spatial consistency between the two SLC single-view complex images, the two SLC single-view complex images are registered first.

[0028] Step 2, obtaining a cross-correlation coefficient matrix related to the to-be-detected pixel in the main image based on a statistical method and an INFLO algorithm, comprising: step 2.1, setting a target window and a matching window: setting a target window with the to-be-detected pixel as the center in the main image, and setting a matching window with the same size as the target window and with the to-be-detected pixel as the center in the auxiliary image.

[0029] Step 2.2, oversampling the SLC pixels in the target window and the matching window respectively, and then calculating the SAR intensity value based on the oversampled SLC pixel values, referring to Figure 2 For the target window and the matching window at this time: the two windows participating in the calculation of the cross-correlation coefficient matrix are called cross-correlation windows; (1) identifying candidate intensity value abnormal pixels in the cross-correlation windows by using a statistical method; (2) identifying intensity value abnormal pixels by using the INFLO algorithm based on the candidate intensity value abnormal pixels, and replacing the intensity values of the intensity value abnormal pixels with the intensity median values in the neighborhood window; (3) calculating the normalized cross-correlation coefficient matrix in the cross-correlation window; preferably, in step 2.2, oversampling each pixel in the cross-correlation window before identifying the candidate intensity value abnormal pixels, so as to further improve the accuracy of the SAR single-view complex image offset calculation result.

[0030] Preferably, in step 2, the statistical method is the Z-score method. The Z-score algorithm is used to obtain the standardized value Z of the intensity value corresponding to each pixel within the cross-correlation window. Pixels with a standardized value Z greater than a preset value are set as candidate intensity value anomalous pixels. The formula for calculating the standardized value Z of the intensity value of each pixel is as follows: , where x is the intensity value of the pixel within the cross-correlation window, μ is the mean intensity value of all pixels within the cross-correlation window, and σ is the standard deviation of the intensity values ​​of all pixels within the cross-correlation window. The Z-score method is used to standardize the intensity values ​​within the cross-correlation window, which can quickly filter out candidate pixels with abnormal intensity.

[0031] Preferably, refer to Figure 3 In step 2.2, based on the candidate intensity value anomalous pixel, the INFLO algorithm is used to identify the intensity value anomalous pixel and replace the intensity value of the intensity value anomalous pixel with the median intensity value in its neighborhood window. Specifically: (1) Calculate the forward k nearest neighbors of the candidate intensity value anomalous pixel: For each candidate intensity value anomalous pixel p, the k nearest neighbors method is used to find its k nearest neighbors. The set of all k nearest neighbors of the candidate intensity value anomalous pixel p is its forward k nearest neighbors, denoted as N. k (p); (2) Calculate the inverse k-nearest neighbor of candidate intensity value anomalous pixels: through the forward k-nearest neighbor N of each candidate intensity value anomalous pixel p. k (p) Calculate the inverse k-nearest neighbor region for each candidate intensity value anomalous pixel p, denoted as RN. k (p); (3) Calculate the INFLO value of candidate intensity value anomalous pixel p and identify intensity value anomalous pixels: The union of the forward k nearest neighbors and the reverse k nearest neighbors of each candidate intensity value anomalous pixel p constitutes the influence space SN of each candidate intensity value anomalous pixel p. k (p), SN k (p)=N k (p)∪RN k (p), calculate the INFLO value of each candidate intensity value anomalous pixel p by influence space, and the candidate intensity value anomalous pixel p with an INFLO value greater than a preset threshold is identified as an intensity value anomalous pixel; (4) Replace the intensity value of the intensity value anomalous pixel: perform the next operation for each intensity value anomalous pixel, take a 7×7 neighborhood window centered on the intensity value anomalous pixel, and replace its intensity value with the median intensity value in this neighborhood window.

[0032] Specifically, the INFLO value calculation process for each candidate intensity value anomalous pixel p is as follows.

[0033] (1) Calculate the reachable density(p) of each candidate intensity value anomalous pixel p: , where k distance(p) Let be the distance to the k-th nearest neighbor of p.

[0034] (2) Calculate the INFLO value of each candidate intensity value abnormal pixel p: , avg density (SN k (p)) is the average value of the reachable density of all pixels in SN k (p), and INFLO(p) is the INFLO value of the candidate intensity value abnormal pixel p.

[0035] Further density analysis is performed on the candidate intensity value abnormal pixels by the INFLO method, which combines the spatial distribution structure information of the pixels, can effectively identify and eliminate the intensity value abnormal pixels with sparse distribution, effectively remove the patch error, and at the same time, retain the pixels with high density and belonging to the image structure edge, thereby improving the problem of discarding the intensity value normal pixels in the existing method. Secondly, the method does not need to test the scale parameter of the intensity distribution in advance, and has better adaptability.

[0036] Specifically, in step 2.2, the normalized cross-correlation coefficient matrix in the cross-correlation window is calculated, specifically: the SAR intensity value is converted from the spatial domain to the frequency domain by Fourier transform, the cross-correlation coefficient of each position in the cross-correlation window is calculated, and the normalized cross-correlation coefficient matrix is obtained.

[0037] Step 3, combine the peak position of the normalized cross-correlation coefficient matrix and the SAR single-view complex image pixel to calculate the offset amount in the range and azimuth directions.

[0038] Specifically, step 3 is specifically: taking the position of the matrix where the peak value is located in the cross-correlation coefficient matrix, the offset amount of this position from the center position of the cross-correlation coefficient matrix is the integral pixel offset amount of the pixel to be tested; according to the cross-correlation coefficient matrix, the correlation coefficient surface fitting is carried out, and the sub-pixel offset amount of the pixel to be tested in the main image can be obtained by solving the extreme value of the fitting function; the integral pixel offset amount and the sub-pixel offset amount are integrated to obtain the final offset amount in the range and azimuth directions.

[0039] Step 4, steps 2-3 are performed pixel by pixel, and the complete ground surface offset amount map covering the regions of the two SAR single-view complex images can be obtained.

[0040] First, the SAR image intensity value is logarithmically transformed and Z-score standardized, the candidate intensity value abnormal pixels are quickly screened out and their position information is recorded, then the INFLO algorithm is used to identify the intensity value abnormal pixels, and the true edge structure information is retained, the accuracy of the intensity value abnormal pixel detection is improved, and the patch error in the SAR image offset tracking is reduced, and at the same time, the loss of normal offset amount is effectively reduced.

[0041] Figure 4 and Figure 5The results of the distance offset of SAR intensity images in a certain area by the traditional SAR single-view complex image offset tracking algorithm and the algorithm of the present application respectively. Figure 4 As can be clearly seen from the results of the distance offset, there are a large number of patch regions (for example, the regions shown by the red boxes). According to the continuity feature of the glacier movement, the regular patch-like results are far from the actual displacement of the glacier surface, and therefore, the patch-like results are not the real distance offset of the glacier, but the matching error caused by the local existence of the abnormal intensity value pixels in the SAR single-view complex image. In contrast, the patch-like part of the distance offset results obtained by the method of the present application is greatly reduced, while the originally normal offset values are retained. The glacier offset results obtained by the method of the present application conform to the continuity feature of the glacier surface movement, and the results are obviously more accurate.

[0042] The above is a further detailed description of the present application in combination with the specific preferred embodiments, and the specific implementation of the present application cannot be limited to these descriptions. For ordinary skilled persons in the technical field to which the present application belongs, a number of simple deductions and replacements can be made without departing from the concept of the present application, and all of them shall be regarded as falling within the protection scope of the present application.

Claims

1. A SAR image offset tracking method that is not affected by patch errors, characterized by, The method comprises the following steps: Step 1, data preprocessing: obtaining two SLC SAR single-view complex images of the same coverage area, and setting them as a main image and an auxiliary image respectively; registering the main image and the auxiliary image based on external DEM, and resampling the auxiliary image to the coordinate system of the main image; Step 2, obtaining a cross-correlation coefficient matrix related to a to-be-detected pixel in the main image based on a statistical method and an INFLO algorithm, comprising: Step 2.1, setting a target window and a matching window: setting a target window with the to-be-detected pixel as the center in the main image, and setting a matching window with the same size as the target window and with the to-be-detected pixel as the center in the auxiliary image; Step 2.2, oversampling SLC pixels in the target window and the matching window respectively, and then calculating SAR intensity values based on the oversampled SLC pixel values, for the target window and the matching window at this time: The two windows participating in the calculation of the cross-correlation coefficient matrix are called cross-correlation windows; (1) identifying candidate intensity value abnormal pixels in the cross-correlation window by using a statistical method; (2) identifying intensity value abnormal pixels based on the candidate intensity value abnormal pixels by using the INFLO algorithm, and replacing the intensity values of the intensity value abnormal pixels with the intensity median values in the neighborhood window of the intensity value abnormal pixels; (3) calculating a normalized cross-correlation coefficient matrix in the cross-correlation window; Step 3, calculating the offset in the range direction and the azimuth direction by using the peak position of the normalized cross-correlation coefficient matrix and the SAR single-view complex image pixel; Step 4, performing steps 2-3 pixel by pixel, so as to obtain a ground surface offset map covering the area of the two SAR single-view complex images.

2. The SAR image offset tracking method without patch error impact according to claim 1, characterized in that, In step 2.2, the INFLO algorithm is used to identify intensity value abnormal pixels based on the candidate intensity value abnormal pixels, and the intensity values of the intensity value abnormal pixels are replaced with the intensity median values in the neighborhood window of the intensity value abnormal pixels, specifically as follows: (1) Calculate the forward k-neighborhood of the candidate intensity value abnormal pixel: for each candidate intensity value abnormal pixel p, find its k nearest neighbors using the k-neighbor method, and the set of all k nearest neighbors of the candidate intensity value abnormal pixel p is its forward k-neighborhood, denoted as N k (p); (2) Computing the reverse k-neighborhood of the candidate intensity value anomaly pixel: by the forward k-neighborhood N k (p) of each candidate intensity value anomaly pixel p k (p) Computing the reverse k-neighborhood RN of each candidate intensity value anomaly pixel p (3) calculating the INFLO value of the candidate intensity value abnormal pixel p and identifying the intensity value abnormal pixel: the union of the forward k-neighbor domain and the reverse k-neighbor domain of each candidate intensity value abnormal pixel p constitutes the influence space SN of each candidate intensity value abnormal pixel p k (p), SN k (p) = N k (p) ∪ RN k (p), the INFLO value of each candidate intensity value abnormal pixel p is calculated through the influence space, and the candidate intensity value abnormal pixel p whose INFLO value is greater than the preset INFLO threshold value is the intensity value abnormal pixel; (4) replacing the intensity values of the intensity value abnormal pixels: taking a 7*7 neighborhood window centered on each intensity value abnormal pixel, and replacing the intensity value of the intensity value abnormal pixel with the intensity median value in the neighborhood window.

3. The SAR image offset tracking method without patch error impact according to claim 2, characterized in that, In step 2, the statistical method is a Z-score method, and a standardized value Z of each pixel in the cross-correlation window is obtained by a Z-score algorithm, and a pixel with a standardized value Z greater than a preset threshold is set as a candidate intensity abnormal pixel, and a calculation formula of the standardized value Z of each pixel is as follows: Wherein x is an intensity value of a pixel in the cross-correlation window, μ is a mean value of intensity values of all pixels in the cross-correlation window, and σ is a standard deviation of the intensity values of all pixels in the cross-correlation window.

4. The SAR image offset tracking method without patch error impact according to claim 3, characterized in that, The INFLO value calculation process of each candidate intensity value abnormal pixel p is as follows: (1) Calculate the reachable density of each candidate intensity value abnormal pixel p: where k distance(p) is the distance of the kth nearest neighbor of p; (2) Calculate the INFLO value of each candidate intensity value abnormal pixel p: , avg density (SN k (p)) is the average value of the reachable density of all pixels in SN k (p), and INFLO(p) is the INFLO value of the candidate intensity value abnormal pixel p.

5. The SAR image offset tracking method without patch error impact according to claim 4, characterized in that, The step 3 is specifically as follows: taking the position of the peak in the matrix in the cross-correlation coefficient matrix, and the offset of the position from the center position of the cross-correlation coefficient matrix is the integer pixel offset of the to-be-detected pixel; performing correlation coefficient surface fitting according to the cross-correlation coefficient matrix, and obtaining the sub-pixel offset of the to-be-detected pixel in the main image by solving the extreme value of the fitting function; and combining the integer pixel offset and the sub-pixel offset to obtain the offsets in the range direction and the azimuth direction. 6.The SAR image shift tracking method without patch error impact according to claim 5, wherein, In step 1, two SAR single-view complex images are registered, and the method is as follows: a conversion relationship lookup table between SAR coordinates and geographic coordinates is generated by using the orbit parameters and imaging geometry parameters of the two SAR single-view complex images and combining the external DEM in the coverage area of the SAR single-view complex images; the external DEM in the coverage area of the SAR single-view complex images is converted to the SAR coordinate system through the lookup table; and the primary and secondary images are registered, and the secondary image is resampled to the primary image coordinate system so as to keep consistent with the primary image in the geographic space.

7. The SAR image offset tracking method without patch error impact according to claim 6, characterized in that, In step 2.2, the normalized cross-correlation coefficient matrix in the cross-correlation window is calculated, and specifically, the SAR intensity value is converted from the spatial domain to the frequency domain through Fourier transform, the cross-correlation coefficient of each position in the cross-correlation window is calculated, and the normalized cross-correlation coefficient matrix is obtained.

Images