A SAR image registration method, apparatus, device and medium

Abstract

This application discloses a SAR image registration method, apparatus, device, and medium, relating to the field of image processing technology. The method includes: acquiring a current SAR image and corresponding historical SAR images; calculating the overlap rate of the current SAR image and historical SAR images according to preset rules, and determining the overlapping region when the overlap rate is not lower than a first preset threshold; selecting a first region image and a second region image from the current SAR image and historical SAR images based on the overlapping region; registering the first region image using a correlation matching method based on the second region image to obtain an initial registered image; determining whether the corresponding correlation coefficient is greater than a second preset threshold; if so, dividing the initial registered image and the second region image into blocks; and registering each block of the initial registered image using a correlation matching method based on each block of the second region image to obtain a final registered image. This method can improve the registration accuracy and speed of SAR images.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a SAR image registration method, apparatus, device, and medium. Background Technology

[0002] Unmanned aerial vehicle (UAV) remote sensing platforms possess advantages such as long endurance, real-time image transmission, high-risk area detection, low cost, and high mobility, thus becoming a powerful complement to satellite remote sensing and manned aerial remote sensing. Synthetic Aperture Radar (SAR) is characterized by all-weather, 24 / 7 operation and excellent detection and differentiation capabilities against man-made and camouflaged targets, providing a more effective means of safeguarding national security. Currently, UAV remote sensing platforms equipped with SAR are widely used in disaster monitoring, topographic mapping, and national defense, playing a significant role. Image registration is widely used in change detection, image fusion, and matching navigation technologies. The accuracy and timeliness of image registration directly affect the accuracy and timeliness of subsequent applications such as change detection and image fusion.

[0003] For UAV-borne SAR systems, characterized by multiple operating modes, unpredictable flight paths, variable imaging angles, and the need for real-time image processing, the corresponding image registration algorithm must be able to quickly and automatically match SAR images with different resolutions, azimuths, incident angles, and effective ranges. Current SAR image registration methods fall into two categories. The first category is feature-based matching methods, specifically point features, line features, and region features. Feature-based matching methods have been widely used in optical image matching; for example, SIFT based on point features and its improved methods have mature applications in optical image retrieval and biometric recognition. Compared to optical images, SAR images are microwave images containing speckle noise, and SAR images formed at different azimuths and incident angles show significant differences in geometric and grayscale features. Therefore, the extracted point features have poor consistency and are prone to mismatches. The second category is methods based on image grayscale information matching, specifically correlation matching and mutual information matching. Grayscale-based image registration methods are suitable for small images where there is only translation or slight rotation between the two images. Their advantage is high registration accuracy, but their disadvantages include high computational cost and sensitivity to changes in the geometric features of the images. Rotation, scaling, or changes in the geometric structure between the two images can easily lead to registration failure. In other words, current SAR image registration techniques are often only applicable to a specific type of image, lacking broad applicability and timeliness.

[0004] In summary, improving the registration accuracy and timeliness of SAR images is a problem that needs to be solved. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a SAR image registration method, apparatus, device, and medium that can improve the registration accuracy and timeliness of SAR images. The specific solution is as follows:

[0006] Firstly, this application discloses a SAR image registration method, including:

[0007] Acquire the current SAR image sent by the UAV, and match the corresponding historical SAR image from the historical image database based on the geographic coordinate information of the current SAR image;

[0008] The overlap rate between the current SAR image and the historical SAR image is calculated according to a preset rule, and it is determined whether the overlap rate is not lower than a first preset threshold.

[0009] If so, the overlapping region is determined, and a first region image and a second region image are selected from the current SAR image and the historical SAR image respectively based on the overlapping region;

[0010] Based on the second region image, the first region image is registered using a correlation matching method to obtain an initial registered image corresponding to the current SAR image, and it is determined whether the corresponding correlation coefficient is greater than a second preset threshold.

[0011] If the value is greater than the given value, the initial registration image and the historical SAR image are processed into blocks, and based on each block in the historical SAR image, the corresponding block in the initial registration image is registered using a correlation matching method to obtain the final registration image.

[0012] Optionally, before calculating the overlap rate between the current SAR image and the historical SAR image according to a preset rule, the method further includes:

[0013] The file headers of the current SAR image and the historical SAR image are read and parsed respectively to obtain the first pixel scale of the current SAR image and the second pixel scale of the historical SAR image;

[0014] Determine whether the first pixel scale and the second pixel scale are consistent. If they are inconsistent, determine the smaller value between the first pixel scale and the second pixel scale, and perform downsampling processing on the SAR image corresponding to the smaller value so that the first pixel scale and the second pixel scale are consistent.

[0015] Optionally, calculating the overlap rate between the current SAR image and the historical SAR image according to a preset rule includes:

[0016] Obtain the first geographical extent of the current SAR image and the second geographical extent of the historical SAR image;

[0017] Based on the first geographic range and the second geographic range, the current SAR image and the historical SAR image are stitched together to obtain a stitched image, and the first geographic coordinates and the second geographic coordinates of the current SAR image and the historical SAR image in the stitched image are determined.

[0018] The overlap rate between the current SAR image and the historical SAR image is calculated based on the first geographic coordinates and the second geographic coordinates.

[0019] Optionally, obtaining the first geographic range of the current SAR image and the second geographic range of the historical SAR image includes:

[0020] The latitude and longitude coordinates, pixel scale, and image width and height of the upper left corner of the current SAR image and the historical SAR image are obtained respectively, and the latitude and longitude coordinates of the lower right corner are determined based on the latitude and longitude coordinates of the upper left corner, the pixel scale, and the image width and height.

[0021] The first geographical extent of the current SAR image and the second geographical extent of the historical SAR image are determined based on the latitude and longitude coordinates of the upper left corner and the lower right corner.

[0022] Optionally, the step of registering the first region image with the second region image using a correlation matching method to obtain an initial registered image corresponding to the current SAR image includes:

[0023] A rectangular data block of a first preset image size is extracted from the second region image based on the center point of the second region image and used as a reference image; and a rectangular data block of a second preset image size is extracted from the first region image based on the center point of the first region image and used as a template image.

[0024] The template image is roamed on the baseline image, and a first correlation coefficient is calculated for each roaming position. Then, the largest correlation coefficient is selected from all the first correlation coefficients of the roaming positions, and the roaming position corresponding to the largest correlation coefficient is determined as the best matching position.

[0025] Based on the optimal matching position, the offset relationship between the current SAR image and the historical SAR image is determined, and the offset relationship is used to perform image transformation on the current SAR image to obtain the corresponding initial registration image.

[0026] Optionally, the step of registering each corresponding block in the initial registration image based on each block in the historical SAR image using a correlation matching method to obtain the final registration image includes:

[0027] Based on each block in the historical SAR image, determine each corresponding block in the initial registration image, and calculate the coordinates of the best matching point between the two blocks;

[0028] Determine whether the second correlation coefficient corresponding to the coordinates of the best matching point is not lower than a third preset threshold. If it is not lower, then the coordinates of the best matching point are used as the control point coordinates of the corresponding block; if it is lower, then the center point coordinates of the corresponding block are used as the control point coordinates.

[0029] A first set of control points is obtained based on the coordinates of all control points in the blocks. A first affine transformation relationship between the initial registration image and the historical SAR image is determined based on the first set of control points. Then, the initial registration image is transformed using the first affine transformation relationship to obtain the final registration image.

[0030] Optionally, after determining whether the corresponding correlation coefficient is greater than the second preset threshold, the method further includes:

[0031] If the value is not greater than the historical SAR image, the current SAR image is registered using the SIFT feature matching method to obtain the final registered image. The process of registering the first region image using the SIFT feature matching method based on the second region image to obtain the final registered image includes: performing filtering and enhancement processing on the first region image and the second region image respectively to obtain corresponding gradient images, acquiring the gradient magnitude map of the gradient images, and constructing a Harris scale space based on the gradient magnitude map; performing extreme point detection on the Harris scale space at different scales, calculating the gradient histogram of each extreme point, and determining whether the corresponding extreme point is a key point based on the gradient histogram; if it is a key point, generating a feature descriptor based on the key point, and performing key point matching using the nearest neighbor distance ratio method based on the feature descriptor to filter out a set of matching points from the historical SAR image; constructing a second control point set based on the set of matching points, determining a second affine transformation relationship between the current SAR image and the historical SAR image based on the second control point set, and then performing image transformation on the current SAR image using the second affine transformation relationship to obtain the final registered image.

[0032] Secondly, this application discloses a SAR image registration device, comprising:

[0033] The SAR image acquisition module is used to acquire the current SAR image sent by the UAV and match the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image.

[0034] The overlap rate calculation module is used to calculate the overlap rate between the current SAR image and the historical SAR image according to a preset rule, and to determine whether the overlap rate is not lower than a first preset threshold.

[0035] The regional image filtering module is used to determine the overlapping region if the condition is met, and to filter out the first regional image and the second regional image from the current SAR image and the historical SAR image respectively based on the overlapping region.

[0036] The initial registration module is used to register the first region image with the second region image using a correlation matching method to obtain an initial registration image corresponding to the current SAR image, and to determine whether the corresponding correlation coefficient is greater than a second preset threshold.

[0037] The final registration module is used to divide the initial registration image and the historical SAR image into blocks if the value is greater than the value, and to register each corresponding block in the initial registration image based on each block in the historical SAR image using a correlation matching method to obtain the final registration image.

[0038] Thirdly, this application discloses an electronic device, including:

[0039] Memory, used to store computer programs;

[0040] A processor is configured to execute the computer program to implement the steps of the aforementioned disclosed SAR image registration method.

[0041] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the steps of the aforementioned disclosed SAR image registration method.

[0042] As can be seen, this application acquires the current SAR image sent by the UAV and matches the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image; calculates the overlap rate of the current SAR image and the historical SAR image according to preset rules, and determines whether the overlap rate is not lower than a first preset threshold; if so, the overlapping area is determined, and a first area image and a second area image are selected from the current SAR image and the historical SAR image based on the overlapping area; the first area image is registered using a correlation matching method based on the second area image to obtain an initial registered image corresponding to the current SAR image, and determines whether the corresponding correlation coefficient is greater than a second preset threshold; if it is greater, the initial registered image and the historical SAR image are divided into blocks, and each block in the historical SAR image is registered using a correlation matching method to obtain a final registered image. Therefore, after acquiring the current SAR image sent by the UAV, this application obtains the corresponding historical SAR image based on the geographic coordinate information, calculates the overlap rate between the current SAR image and the historical SAR image, and, when the overlap rate is not lower than a first preset threshold, filters out a first region image and a second region image from the current SAR image and the historical SAR image respectively based on the overlapping region. Then, it uses a correlation matching method to register the first region image and the second region image to obtain an initial registered image of the current SAR image relative to the historical SAR image, and determines whether the correlation coefficient at this time is greater than the second preset threshold. If it is, it further divides the initial registered image and the historical SAR image into blocks, and uses the correlation matching method again to register each block of the initial registered image and the historical SAR image to obtain the final registered image of the initial registered image relative to the historical SAR image. In this way, this application transforms the registration problem between SAR images into the registration problem between overlapping regions, which can eliminate invalid feature points, thereby improving the timeliness and accuracy of image registration; furthermore, this application also improves the accuracy of image registration through two matching processes, that is, by adopting a matching strategy from coarse matching to fine matching. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0044] Figure 1 This is a flowchart of a SAR image registration method disclosed in this application;

[0045] Figure 2 This is a schematic diagram of a specific SAR image registration process disclosed in this application;

[0046] Figure 3 This is a flowchart of a specific SAR image registration method disclosed in this application;

[0047] Figure 4 This is a flowchart of a specific SAR image registration method disclosed in this application;

[0048] Figure 5 (a) is a specific simulated SAR image disclosed in this application;

[0049] Figure 5 (b) is a gradient magnitude map generated after ROEWA filtering as disclosed in this application;

[0050] Figure 5 (c) is a gradient direction pattern generated after ROEWA filtering as disclosed in this application;

[0051] Figure 6 This is a current SAR image of a test dataset I disclosed in this application;

[0052] Figure 7 This application discloses a test dataset I containing historical SAR images.

[0053] Figure 8 This is a schematic diagram of the image registration results for a test dataset I disclosed in this application;

[0054] Figure 9 This is a current SAR image of a test dataset II disclosed in this application;

[0055] Figure 10 This application discloses a test dataset II of historical SAR images.

[0056] Figure 11 This is a schematic diagram of the image registration results for a test dataset II disclosed in this application;

[0057] Figure 12 This is a schematic diagram of the structure of a SAR image registration device disclosed in this application;

[0058] Figure 13 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0059] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0060] Current SAR image registration methods fall into two categories. The first is feature-matching, which is widely used in optical image matching. However, compared to optical images, SAR images are microwave imaging, containing speckle noise. Furthermore, SAR images formed at different azimuths and incident angles exhibit significant differences in geometric and grayscale features, resulting in poor consistency of extracted point features and a high risk of mismatches. The second category is grayscale-based matching, suitable for smaller images where there is only translation or slight rotation between the two images. Its advantages include high registration accuracy, but its disadvantages include high computational cost and sensitivity to changes in image geometry. Rotation, scaling, or changes in geometric structure between the two images can easily lead to registration failure. In other words, current SAR image registration techniques are often only applicable to a specific type of image, lacking broad applicability and timeliness. Therefore, this application discloses a SAR image registration method, apparatus, device, and medium that can improve the registration accuracy and timeliness of SAR images.

[0061] See Figure 1 and Figure 2 As shown in the figure, this application discloses a SAR image registration method, which includes:

[0062] Step S11: Obtain the current SAR image sent by the UAV, and match the corresponding historical SAR image from the historical image database based on the geographic coordinate information of the current SAR image.

[0063] In this embodiment, when the UAV performs a reconnaissance mission, it transmits SAR images it has captured. It should be noted that the current SAR image transmitted by the UAV is a GeoTiff (GeoTag ImageFile Format, i.e., a geospatial raster data storage format based on TIFF) image that has undergone system-level geometric correction. GeoTiff is a common remote sensing image data format; it is raster data stored as a matrix of pixels. Furthermore, the image contains geographic coordinate information, and historical images are retrieved from the historical image database based on the geographic coordinate information and time conditions of the current SAR image to obtain historical SAR images of the same area as the current SAR image. These historical SAR images are also in GeoTiff format, thus completing the matching of the two images. Since the two images, after system-level geometric correction, only have translation and a small rotation angle, the geographic information can be used to narrow down the image matching range and improve matching accuracy, while also increasing matching speed, which is beneficial for real-time image registration.

[0064] Step S12: Calculate the overlap rate of the current SAR image and the historical SAR image according to the preset rules, and determine whether the overlap rate is not lower than the first preset threshold.

[0065] In this embodiment, it is necessary to calculate the overlap rate between the current SAR image and the historical SAR image according to a preset rule, and determine whether the overlap rate is not lower than a first preset threshold. The first preset threshold can be set to 30%.

[0066] It should be noted that before calculating the overlap rate of the current SAR image and the historical SAR image according to the preset rules, the process further includes: reading and parsing the file headers of the current SAR image and the historical SAR image respectively to obtain the first pixel scale of the current SAR image and the second pixel scale of the historical SAR image; determining whether the first pixel scale and the second pixel scale are consistent; if they are inconsistent, determining the smaller value of the first pixel scale and the second pixel scale, and downsampling the SAR image corresponding to the smaller value so that the first pixel scale and the second pixel scale are consistent. It is understood that before calculating the overlap rate of the current SAR image and the historical SAR image, it is also necessary to determine whether the pixel scales of the current SAR image and the historical SAR image are consistent. In this embodiment, the scale consistency of the two images needs to be satisfied to improve the success rate of subsequent image matching. Specifically, the pixel scale parameters in the headers of the current and historical SAR images are parsed to obtain the first pixel scale of the current SAR image and the second pixel scale of the historical SAR image. The pixel scale is one of the attributes of the Geotiff file, representing the pixel size in "degrees". Then, it is determined whether the first and second pixel scales are consistent. If they are inconsistent, the SAR image with the smaller scale value needs to be downsampled to ensure that the pixel scales of the two images are consistent after downsampling. For example, if the current SAR image and the historical SAR image are denoted as M1 and M2 respectively, and assuming that the first pixel scale of M1 is s1=0.0005 and the second pixel scale of M2 is s2=0.0003, then M2 needs to be downsampled. The downsampling scale factor is F=s1 / s2. Specifically, bilinear interpolation can be used to downsample the image. After downsampling, M1 and M2 have scale consistency, which is beneficial for subsequent image matching.

[0067] Step S13: If yes, then the overlapping region is determined, and based on the overlapping region, the first region image and the second region image are selected from the current SAR image and the historical SAR image, respectively.

[0068] In one specific embodiment, if the overlap rate is not lower than a first preset threshold, the rectangular pixel coordinates corresponding to the overlapping region in the current SAR image and the historical SAR image are determined, and the first region image and the second region image are selected from the current SAR image and the historical SAR image based on the rectangular pixel coordinates.

[0069] In another specific embodiment, if the overlap rate is lower than a first preset threshold, a prompt message "Image overlap rate does not meet registration requirements" is given.

[0070] Step S14: Based on the second region image, the first region image is registered using a correlation matching method to obtain an initial registered image corresponding to the current SAR image, and it is determined whether the corresponding correlation coefficient is greater than a second preset threshold.

[0071] In this embodiment, a correlation matching method is used to register the first region image and the second region image to obtain an initial registration image of the current SAR image relative to historical SAR images. This initial registration image represents the coarse matching position between the current SAR image and the historical SAR image. It is then determined whether the correlation coefficient corresponding to the coarse matching position is greater than a second preset threshold, where the second preset threshold can be set to 0.5.

[0072] Step S15: If the value is greater than the given value, the initial registration image and the historical SAR image are processed into blocks respectively. Based on each block in the historical SAR image, the corresponding block in the initial registration image is registered using a correlation matching method to obtain the final registration image.

[0073] In this embodiment, if the correlation coefficient is greater than the second preset threshold, the overlapping area data, i.e., the initial registration image and the historical SAR image, are further refined by a correlation matching method to complete the image registration. Specifically, the initial registration image and the historical SAR image are divided into blocks, and the correlation matching method is used again to register each block of the initial registration image and the historical SAR image to obtain the final registration image of the initial registration image relative to the historical SAR image.

[0074] As can be seen, this application acquires the current SAR image sent by the UAV and matches the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image; calculates the overlap rate of the current SAR image and the historical SAR image according to preset rules, and determines whether the overlap rate is not lower than a first preset threshold; if so, the overlapping area is determined, and a first area image and a second area image are selected from the current SAR image and the historical SAR image based on the overlapping area; the first area image is registered using a correlation matching method based on the second area image to obtain an initial registered image corresponding to the current SAR image, and determines whether the corresponding correlation coefficient is greater than a second preset threshold; if it is greater, the initial registered image and the historical SAR image are divided into blocks, and each block in the historical SAR image is registered using a correlation matching method to obtain a final registered image. Therefore, after acquiring the current SAR image sent by the UAV, this application obtains the corresponding historical SAR image based on the geographic coordinate information, calculates the overlap rate between the current SAR image and the historical SAR image, and, when the overlap rate is not lower than a first preset threshold, filters out a first region image and a second region image from the current SAR image and the historical SAR image respectively based on the overlapping region. Then, it uses a correlation matching method to register the first region image and the second region image to obtain an initial registered image of the current SAR image relative to the historical SAR image, and determines whether the correlation coefficient at this time is greater than the second preset threshold. If it is, it further divides the initial registered image and the historical SAR image into blocks, and uses the correlation matching method again to register each block of the initial registered image and the historical SAR image to obtain the final registered image of the initial registered image relative to the historical SAR image. In this way, this application transforms the registration problem between SAR images into the registration problem between overlapping regions, which can eliminate invalid feature points, thereby improving the timeliness and accuracy of image registration; furthermore, this application also improves the accuracy of image registration through two matching processes, that is, by adopting a matching strategy from coarse matching to fine matching.

[0075] See Figure 3 As shown, this application discloses a specific SAR image registration method. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution. Specifically, it includes:

[0076] Step S21: Obtain the current SAR image sent by the UAV, and match the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image.

[0077] Step S22: Calculate the overlap rate of the current SAR image and the historical SAR image according to the preset rules, and determine whether the overlap rate is not lower than the first preset threshold.

[0078] Step S23: If yes, then the overlapping region is determined, and based on the overlapping region, the first region image and the second region image are selected from the current SAR image and the historical SAR image, respectively.

[0079] Step S24: Based on the center point of the second region image, extract a rectangular data block of the first preset image size from the second region image and use it as a reference image; and based on the center point of the first region image, extract a rectangular data block of the second preset image size from the first region image and use it as a template image.

[0080] In this embodiment, rectangular data blocks of a first preset image size and a second preset image size are extracted from the second region image and the first region image, respectively, based on the center points of the second region image and the first region image, and used as the reference image and the template image, respectively. For example, in this embodiment, the second region image can be denoted as I2 and the first region image can be denoted as I1. Then, using the center point of I2 as the center point of the reference image, a rectangular data block of size 384×384 is extracted as the matching reference image; using the center point of I1 as the center point of the template image, a rectangular data block of size 256×256 is extracted as the matching template image.

[0081] Step S25: The template image is roamed on the reference image, and the first correlation coefficient of each roaming position is calculated. Then, the maximum correlation coefficient is selected from all the first correlation coefficients of the roaming positions, and the roaming position corresponding to the maximum correlation coefficient is determined as the best matching position.

[0082] In this embodiment, the template image is roamed on the baseline image, and a first correlation coefficient is calculated for each roaming position. Then, the position with the highest correlation coefficient is selected from all the first correlation coefficients of the roaming positions, and the roaming position corresponding to the highest correlation coefficient is determined as the best matching position. That is, the template image roams on the baseline image and the correlation coefficient is calculated, and the position with the highest correlation coefficient is taken as the best matching position. The correlation coefficient is calculated as follows:

[0083] ;

[0084] in, Represents the correlation coefficient. Indicates the template image in The pixel grayscale value at the point. This represents the mean of the template image. Indicates the reference image in The pixel grayscale value at the point. This indicates the roaming position of the template image within the reference image. Indicates the reference image in displacement The mean value at point M and N represent the number of vertical and horizontal pixels in the template image, respectively.

[0085] Step S26: Determine the offset relationship between the current SAR image and the historical SAR image based on the optimal matching position, and use the offset relationship to perform image transformation on the current SAR image to obtain the corresponding initial registration image, and determine whether the corresponding correlation coefficient is greater than the second preset threshold.

[0086] In this embodiment, the offset relationship between the current SAR image and historical SAR images is determined based on the optimal matching position. This involves converting the coordinates of the optimal matching position into offset coordinates of the current SAR image within the historical SAR image. Then, a translation transformation is performed on the current SAR image based on these offset coordinates to obtain the corresponding initial registration image. This achieves coarse registration between the initial registration image and the historical SAR image. Additionally, it is necessary to determine whether the correlation coefficient corresponding to the optimal matching position is greater than a second preset threshold of 0.5, and to execute subsequent steps based on the determination result.

[0087] Step S27: If the value is greater than the given value, the initial registration image and the historical SAR image are processed into blocks respectively, and based on each block in the historical SAR image, the corresponding block in the initial registration image is registered using the correlation matching method to obtain the final registration image.

[0088] In this embodiment, the above-mentioned registration of each corresponding block in the initial registration image based on each block in the historical SAR image using a correlation matching method to obtain the final registration image includes: determining each corresponding block in the initial registration image based on each block in the historical SAR image, and calculating the best matching point coordinates between two blocks; determining whether the second correlation coefficient corresponding to the best matching point coordinates is not lower than a third preset threshold; if it is not lower, then the best matching point coordinates are used as the control point coordinates of the corresponding block; if it is lower, then the center point coordinates of the corresponding block are used as the control point coordinates; obtaining a first control point set based on the control point coordinates of all blocks, and determining a first affine transformation relationship between the initial registration image and the historical SAR image according to the first control point set; and then using the first affine transformation relationship to perform image transformation on the initial registration image to obtain the final registration image. Based on the obtained initial registration image and historical SAR image, the initial registration image and historical SAR image are divided into blocks to achieve fine registration. The block size can be 128×128. The coordinates of the best matching point for each block are calculated. Then, it is determined whether the second correlation coefficient corresponding to the best matching point coordinates is not lower than a third preset threshold, which can be set to 0.7. If it is not lower than 0.7, the coordinates of the best matching point are used as the control point coordinates of the corresponding block; if it is lower than 0.7, the coordinates of the center point of the block are used as the control point coordinates. The control point coordinates of all blocks are calculated, and the first control point set of the initial registration image registered to the historical SAR image is obtained. The number of control points in the set is equal to the number of blocks. The expression for the i-th control point is as follows:

[0089] ;

[0090] in, and , respectively, are the pixel coordinates of the i-th control point in the initial registration image and the historical SRA image.

[0091] Then, based on the first set of control points, the first affine transformation relationship between the initial registration image and the historical SAR image is determined, which is the affine transformation coefficient. In order to use the first affine transformation relationship to perform an affine transformation on the initial registration image to obtain the transformed image registered to the historical SAR image, which is the final registration image.

[0092] For more detailed processing procedures regarding steps S21, S22, and S23, please refer to the corresponding content disclosed in the foregoing embodiments, which will not be repeated here.

[0093] As can be seen, when using the correlation matching method to register the first region image and the second region image to obtain the initial registered image corresponding to the current SAR image, specifically, based on the center points of the second region image and the first region image, rectangular data blocks of the first preset image size and the second preset image size are extracted from the second region image and the first region image respectively as the reference image and the template image. The template image is roamed on the reference image, and the first correlation coefficient of each roaming position is calculated. Then, the largest correlation coefficient is selected from the first correlation coefficients of all roaming positions, and the roaming position corresponding to the largest correlation coefficient is determined as the best matching position. Finally, the offset coordinates between the current SAR image and the historical SAR image are determined based on the best matching position, and then the current SAR image is translated according to the offset coordinates to obtain the corresponding initial registered image after transformation, thus completing the coarse registration process of the image. Then, when the correlation coefficient corresponding to the optimal matching position is greater than the second preset threshold, fine registration is further performed using a correlation matching method. Specifically, the initial registration image and the historical SAR image are divided into blocks, and the coordinates of the nearest matching point in each block are determined. This leads to the determination of control point coordinates, and the first affine transformation relationship between the initial registration image and the historical SAR image is determined using the first control point set formed by the control point coordinates. Finally, the initial registration image is subjected to an affine transformation using the first affine transformation relationship to obtain the transformed image registered to the historical SAR image, which is the final registration image. In this way, by first performing coarse matching on the overlapping data, and then performing block-based fine matching on the coarse matching results, the accuracy of image registration is improved through two matching processes.

[0094] See Figure 4 As shown, this application discloses a specific SAR image registration method. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution. Specifically, it includes:

[0095] Step S31: Obtain the current SAR image sent by the UAV, and match the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image.

[0096] Step S32: Obtain the first geographical extent of the current SAR image and the second geographical extent of the historical SAR image.

[0097] In this embodiment, obtaining the first geographical extent of the current SAR image and the second geographical extent of the historical SAR image includes: obtaining the latitude and longitude coordinates, pixel scale, and image width and height of the upper left corner of the current SAR image and the historical SAR image respectively, and determining the latitude and longitude coordinates of the lower right corner based on the latitude and longitude coordinates of the upper left corner, the pixel scale, and the image width and height; and determining the first geographical extent of the current SAR image and the second geographical extent of the historical SAR image based on the latitude and longitude coordinates of the upper left corner and the lower right corner. That is, this embodiment first determines the latitude and longitude coordinates of the lower right corner by obtaining the latitude and longitude coordinates, pixel scale, and image width and height of the upper left corner of the current SAR image and the historical SAR image, and then determines the first geographical extent of the current SAR image and the second geographical extent of the historical SAR image based on the latitude and longitude coordinates of the upper left corner and the lower right corner. For example, taking the current SAR image M1 as an example, the latitude and longitude coordinates of the upper left corner of M1 are analyzed. Pixel scale Image width and height Calculate the latitude and longitude coordinates of the lower right corner of M1. :

[0098] ;

[0099] Obtain the first geographic range of M1 , :

[0100]

[0101] Similarly, the latitude and longitude coordinates of the upper left corner of the historical SAR image M2 are obtained. and latitude and longitude coordinates in the lower right corner And the corresponding second geographical range.

[0102] Step S33: Perform image stitching processing on the current SAR image and the historical SAR image based on the first geographical range and the second geographical range to obtain a stitched image, and determine the first geographical coordinates and the second geographical coordinates of the current SAR image and the historical SAR image in the stitched image, and calculate the overlap rate of the current SAR image and the historical SAR image based on the first geographical coordinates and the second geographical coordinates.

[0103] In this embodiment, the current SAR image and historical SAR image are stitched together based on a first geographical range and a second geographical range to obtain a stitched image, which is also the geographical range of the stitched image. , The specific method is as follows:

[0104] 1) If ,but ;

[0105] 2) If ,but ;

[0106] 3) If ,but ;

[0107] 4) If ,but .

[0108] Then, the first and second geographic coordinates of the current SAR image and the historical SAR image in the stitched image are calculated. Specifically, the first pixel coordinates of the upper left and lower right corners of image M1 in the stitched image are calculated, and the second pixel coordinates of the upper left and lower right corners of image M2 in the stitched image are calculated. Based on the first and second geographic coordinates, the overlap rate of the current SAR image and the historical SAR image, as well as the pixel coordinates of the rectangles in M1 and M2 where the overlapping area is located, are calculated.

[0109] Step S34: If yes, then the overlapping region is determined, and based on the overlapping region, the first region image and the second region image are selected from the current SAR image and the historical SAR image, respectively.

[0110] In this embodiment, overlapping area data for matching is extracted from the current SAR image and historical SAR image based on the pixel coordinates of the rectangular boxes located in M1 and M2, to obtain the first region image and the second region image respectively.

[0111] Step S35: Based on the second region image, the first region image is registered using a correlation matching method to obtain an initial registered image corresponding to the current SAR image, and it is determined whether the corresponding correlation coefficient is greater than a second preset threshold.

[0112] Step S36: If it is not greater than, then the current SAR image is registered using the SIFT feature matching method based on the historical SAR image to obtain the final registered image.

[0113] In this embodiment, it can be understood that although the two images are theoretically only translated after geocorrection, due to the existence of positioning errors, there may be a small angle rotation between the current SAR image and the historical SAR image. That is, the corresponding correlation coefficient will be lower than the second preset threshold of 0.5. Therefore, the registration process using the correlation matching method is not reliable.

[0114] The process of registering the first region image with the second region image using the SIFT (Scale Invariant Feature Transform) feature matching method to obtain the final registered image includes: performing filtering and enhancement processing on the first region image and the second region image respectively to obtain corresponding gradient images, and obtaining the gradient magnitude map of the gradient image, and constructing a Harris scale space based on the gradient magnitude map; performing extreme point detection on the Harris scale space at different scales, calculating the gradient histogram of each extreme point, and determining whether the corresponding extreme point is a key point based on the gradient histogram; if it is a key point, generating a feature descriptor based on the key point, and performing key point matching using the nearest neighbor distance ratio method based on the feature descriptor to select a set of matching points from the historical SAR images; constructing a second control point set based on the set of matching points, and determining a second affine transformation relationship between the current SAR image and the historical SAR image according to the second control point set, and then performing image transformation on the current SAR image using the second affine transformation relationship to obtain the final registered image.

[0115] In other words, the overlapping region data, namely the first and second region images, are first filtered and enhanced to reduce the influence of speckle noise in the SAR image. Specifically, the Ratio of Exponentially Weighted Averages (ROEWA) operator can be used to filter the image to obtain an edge-enhanced gradient image. The expressions for the ROEWA filtering operator in the horizontal and vertical directions are as follows:

[0116]

[0117]

[0118] In the formula, This represents the scaling factor, initially set to 2, and varies at different scales. , It is a constant, and its value is... x and y represent the pixel coordinates of the image. and j represent the pixel coordinate indices within the filtering window, respectively. Indicates the length of the filtering window. Indicates the width of the filtering window. Indicates the vertical direction. Indicates the horizontal direction.

[0119] Next, the magnitude and direction of the gradient image are calculated using the following formulas:

[0120]

[0121] In the formula, Represents the magnitude of the gradient image. Indicates the direction of the gradient image. Representing scale The horizontal gradient value at time, Representing scale The vertical gradient value at that time.

[0122] See Figure 5 As shown, Figure 5 This application discloses a simulated SAR image and its gradient magnitude map and direction map generated after being filtered by ROEWA.

[0123] Then, a Harris scale space is constructed based on the generated gradient magnitude map. Compared with the DOG (Difference of Gaussian) scale space in the original SIFT algorithm, the Harris scale space can better describe the linear structure in the image, such as the edges and corners of objects, thus exhibiting better noise resistance. The construction expression for the Harris scale space is as follows, where the first expression indicates that the Hessian matrix formed from the gradient magnitude data is scaled with a parameter of . The Gaussian filter is used for filtering:

[0124]

[0125]

[0126] In the formula, The scale parameter is represented as Gaussian filter at time, These are weighting coefficients. Represents convolution. Representing scale The horizontal gradient value at time, Representing scale The vertical gradient value at that time; This represents the value of the determinant of the matrix. The trace of a matrix is ​​the sum of the elements on its main diagonal. It is an empirical constant, typically with a value of 0.04.

[0127] Furthermore, extreme points are detected in Harris scale spatial images at different scales. Extreme points should meet two conditions: (1) greater than a predefined threshold; (2) greater than the value of the 8-neighborhood. After detecting the initial extreme points, the detected extreme points can also be suppressed by thresholding and non-maximum values, and unstable feature points can be deleted using stability metrics to determine the extreme point set. For each extreme point, a gradient histogram is calculated, and the extreme point is judged as a key point based on the gradient histogram. Specifically, the horizontal axis of the gradient histogram is the quantized gradient direction. The original gradient direction value range is [0-180], and after quantization, its value range is reduced to [0-HIST_BIN], where HIST_BIN is 18. The vertical axis of the gradient histogram is the sum of all gradient magnitudes corresponding to the horizontal axis within the circular region centered on the extreme point, that is, traversing all gradient magnitudes within the circular region. If its corresponding gradient direction, after quantization, falls into a certain BIN interval, for example, bin=5, then ; For the generated Gaussian filtering is performed to eliminate the influence of outliers; the filter coefficients are... For the filtered The peak value is obtained, and it is determined whether there is a local maximum. If there is, the extreme point is a key point. The key point can be used for image matching. Its attribute information includes: the pixel coordinates of the extreme point, scale parameter, gradient direction and gradient magnitude.

[0128] For each keypoint, a feature descriptor is generated. Specifically, the GLOH (Gradient Location-Orientation Histogram) descriptor can be used. GLOH is a log-polar coordinate descriptor, and compared with the square SIFT feature descriptor with 4×4 regions, GLOH has more robust matching performance. Assuming the radius ratio factor of the three circular regions of GLOH is γ, the radii of the three circular regions are R={3γ,4.11γ,12γ}. Starting from the outer edge of the first inner circle, the descriptor is divided into 8 fan-shaped directions, resulting in a total of 17 regions including the inner circle. In each region, the gradient histogram is quantized into 8 directions, resulting in a descriptor with a 17×8=136-dimensional feature vector.

[0129] Based on the generated feature descriptors, the Nearest Neighbor Distance Ratio (NNDR) method is used to match key points. NNDR matching mainly consists of two steps:

[0130] (1) Nearest neighbor search: Find the matching feature point descriptors using the minimum Euclidean distance. The corresponding formula is:

[0131]

[0132] in , , and These are feature point sets for the current SAR image and historical SAR images, respectively. and They are respectively and A certain feature point in, For Euclidean distance, These are feature points in historical SAR images that meet the nearest neighbor criteria.

[0133] (2) Nearest neighbor distance ratio: When the ratio of the nearest neighbor distance to the second nearest neighbor distance is less than a certain threshold, it indicates that the current feature point is a preliminary matching point.

[0134] Finally, the Random Sample Consensus (RANSAC) algorithm is used to remove erroneous matching points, resulting in a set of matching points, which is also the second set of control points that can be used for image transformation. The second set of control points is used to determine the second affine transformation relationship between the current SAR image and historical SAR images, so that the current SAR image can be transformed using the second affine transformation relationship to obtain the final registered image.

[0135] For more detailed processing of steps S31 and S35, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.

[0136] As can be seen, in this embodiment, the specific rule for calculating the overlap rate between the current SAR image and the historical SAR image is to obtain the first geographical range of the current SAR image and the second geographical range of the historical SAR image, respectively, and then perform image stitching processing on the current SAR image and the historical SAR image. Next, the first geographical coordinates and the second geographical coordinates of the current SAR image and the historical SAR image in the stitched image are calculated, and then the overlap rate between the current SAR image and the historical SAR image is calculated based on the first geographical coordinates and the second geographical coordinates. By transforming the overall matching problem of two large-size image data into a matching problem of small-size overlapping data, the collection of invalid feature points in non-overlapping areas can be automatically eliminated, significantly improving the timeliness and accuracy of image registration. Furthermore, although the two images, after geocorrection, theoretically only exhibit translation, positioning errors may cause a small-angle rotation between the current SAR image and the historical SAR image. This means the correlation coefficient will be lower than the second preset threshold of 0.5. Therefore, an improved SIFT feature matching method is used for image matching. If feature matching is successful, the RANSAC algorithm is used to extract control point pairs, and the current SAR image is corrected by affine transformation and registered to the historical SAR image. Moreover, the SAR image is enhanced before feature point detection. The enhanced image has clear gradient and orientation features, which is beneficial for feature point detection and matching. The detected features have good consistency, improving image registration accuracy. In other words, this application combines a grayscale-based matching method with a feature detection-based matching method, improving the success rate of image registration.

[0137] The following uses real-flight SAR data acquired by a certain type of UAV to illustrate the image registration accuracy and operational efficiency after the implementation of this invention.

[0138] 1. Test Dataset I

[0139] The imaging region corresponding to dataset I is an image of the urban area of ​​a prefecture-level city in western my country. After system-level geometric correction, the geometric relationship between the two images is mainly translation and scaling. Due to the influence of positioning error, there is a slight rotation between the images.

[0140] Current SAR image M1 as Figure 6 As shown, the main land cover type in the image is buildings, and the relevant parameters are as follows: (1) Image size: (2) Imaging track angle: 18.2 degrees; (3) Radar range: 35km; (4) Geotiff pixel size: 0.25m.

[0141] Historical SAR image M2 such as Figure 7 As shown, the relevant parameters are as follows: (1) Image size: (2) Imaging track angle: 202.9 degrees; (3) Radar range: 45km; (4) Geotiff pixel size: 0.5m.

[0142] See Figure 8 As shown, Figure 8 This application discloses a fused image of a current SAR image and a historical SAR image after registration. Figure 8 As can be seen, the edges of the two images have good consistency, which indicates that the present invention has high registration accuracy.

[0143] 2. Test Dataset II

[0144] Dataset II corresponds to an image of the suburbs of a prefecture-level city in western my country. After system-level geometric correction, the geometric relationship between the two images is mainly translation. Due to the influence of positioning errors, there is a slight rotation between the images. Compared with test dataset I, dataset II has a more uniform land cover type. Due to the significant difference in imaging azimuth angles, the scattering characteristics of the land cover in the images also differ.

[0145] Current SAR image as Figure 9 As shown, the main land cover type in the image is factory buildings, and the relevant parameters are as follows: (1) Image size: (2) Imaging track angle: 9.8 degrees; (3) Radar range: 40km; (4) Geotiff pixel size: 0.4m.

[0146] Historical SAR images such as Figure 10 As shown, the relevant parameters are as follows: (1) Image size: (2) Imaging track angle: 26.5 degrees; (3) Radar range: 45km; (4) Geotiff pixel size: 0.5m.

[0147] See Figure 11 As shown, Figure 11 This is a fused image of a current image and a historical image after registration, as disclosed in this application. As can be seen from the image, the edges of the two images have good consistency, indicating that the present invention has high registration accuracy.

[0148] Furthermore, this embodiment of the application uses a DELL 7050 computer (CPU: i7-7770 3.6GHz, 8GB RAM) and employs C++ programming to implement the SAR image registration process described in this invention. For the registration of two image data sets in test dataset I, the time taken from file reading to outputting the registered image is 0.45 seconds; the same processing is performed on test dataset II, taking 0.41 seconds; both meet the requirements of real-time processing, demonstrating the high timeliness of this invention.

[0149] See Figure 12 As shown in the figure, this application discloses a SAR image registration device, which includes:

[0150] The SAR image acquisition module 11 is used to acquire the current SAR image sent by the UAV and match the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image.

[0151] The overlap rate calculation module 12 is used to calculate the overlap rate between the current SAR image and the historical SAR image according to a preset rule, and to determine whether the overlap rate is not lower than a first preset threshold.

[0152] The regional image filtering module 13 is used to determine the overlapping region if the condition is met, and to filter out the first regional image and the second regional image from the current SAR image and the historical SAR image respectively based on the overlapping region.

[0153] The initial registration module 14 is used to register the first region image with the second region image using a correlation matching method to obtain an initial registration image corresponding to the current SAR image, and to determine whether the corresponding correlation coefficient is greater than a second preset threshold.

[0154] The final registration module 15 is used to perform block processing on the initial registration image and the historical SAR image respectively if the value is greater than the value, and to register each corresponding block in the initial registration image based on each block in the historical SAR image using a correlation matching method to obtain the final registration image.

[0155] As can be seen, this application acquires the current SAR image sent by the UAV and matches the corresponding historical SAR image from the historical image library based on the geographic coordinate information of the current SAR image; calculates the overlap rate of the current SAR image and the historical SAR image according to preset rules, and determines whether the overlap rate is not lower than a first preset threshold; if so, the overlapping area is determined, and a first area image and a second area image are selected from the current SAR image and the historical SAR image based on the overlapping area; the first area image is registered using a correlation matching method based on the second area image to obtain an initial registered image corresponding to the current SAR image, and determines whether the corresponding correlation coefficient is greater than a second preset threshold; if it is greater, the initial registered image and the historical SAR image are divided into blocks, and each block in the historical SAR image is registered using a correlation matching method to obtain a final registered image. Therefore, after acquiring the current SAR image sent by the UAV, this application obtains the corresponding historical SAR image based on the geographic coordinate information, calculates the overlap rate between the current SAR image and the historical SAR image, and, when the overlap rate is not lower than a first preset threshold, filters out a first region image and a second region image from the current SAR image and the historical SAR image respectively based on the overlapping region. Then, it uses a correlation matching method to register the first region image and the second region image to obtain an initial registered image of the current SAR image relative to the historical SAR image, and determines whether the correlation coefficient at this time is greater than the second preset threshold. If it is, it further divides the initial registered image and the historical SAR image into blocks, and uses the correlation matching method again to register each block of the initial registered image and the historical SAR image to obtain the final registered image of the initial registered image relative to the historical SAR image. In this way, this application transforms the registration problem between SAR images into the registration problem between overlapping regions, which can eliminate invalid feature points, thereby improving the timeliness and accuracy of image registration; furthermore, this application also improves the accuracy of image registration through two matching processes, that is, by adopting a matching strategy from coarse matching to fine matching.

[0156] Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Specifically, it may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the SAR image registration method performed by the electronic device disclosed in any of the foregoing embodiments.

[0157] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0158] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0159] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored on it include operating system 221, computer program 222 and data 223, etc., and the storage method can be temporary storage or permanent storage.

[0160] The operating system 221 manages and controls the various hardware devices and computer programs 222 on the electronic device 20 to enable the processor 21 to perform calculations and processing on the massive amounts of data 223 in the memory 22. The operating system 221 can be Windows, Unix, Linux, etc. The computer program 222, in addition to including a computer program capable of performing the SAR image registration method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include data received by the electronic device from external devices, as well as data collected by its own input / output interface 25.

[0161] Furthermore, embodiments of this application also disclose a computer-readable storage medium storing a computer program. When the computer program is loaded and executed by a processor, it implements the method steps performed during SAR image registration as disclosed in any of the foregoing embodiments.

[0162] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0163] Those skilled in the art will further recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0164] The methods or steps of the methods described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0165] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0166] The SAR image registration method, apparatus, device, and storage medium provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method of SAR image registration, characterized by, The method comprises the following steps: acquiring a current SAR image sent by a UAV, and matching a corresponding historical SAR image from a historical image library based on geographical coordinate information of the current SAR image; calculating an overlap rate of the current SAR image and the historical SAR image according to a preset rule, and determining whether the overlap rate is not lower than a first preset threshold; if yes, determining an overlap region, and screening a first region image and a second region image from the current SAR image and the historical SAR image respectively based on the overlap region; using a correlation matching method based on the second region image to register the first region image to obtain an initial registration image corresponding to the current SAR image, and determining whether a corresponding correlation coefficient is greater than a second preset threshold; if yes, performing block processing on the initial registration image and the historical SAR image respectively, and using a correlation matching method based on each block in the historical SAR image to register each corresponding block in the initial registration image to obtain a final registration image; if no, using a SIFT feature matching method based on the historical SAR image to register the current SAR image to obtain a final registration image; wherein the process of using a SIFT feature matching method based on the second region image to register the first region image to obtain a final registration image comprises: performing filtering enhancement processing on the first region image and the second region image respectively to obtain corresponding gradient images, acquiring a gradient amplitude graph of the gradient images, and constructing a Harris scale space based on the gradient amplitude graph; performing extreme point detection on the Harris scale space under different scales, calculating a gradient histogram of each extreme point, and determining whether the corresponding extreme point is a key point based on the gradient histogram; if it is a key point, generating a feature descriptor based on the key point, and using a nearest neighbor distance ratio method based on the feature descriptor to perform key point matching to screen a matching point set from the historical SAR image; constructing a second control point set based on the matching point set, determining a second affine transformation relationship between the current SAR image and the historical SAR image according to the second control point set, and then performing image transformation on the current SAR image using the second affine transformation relationship to obtain a final registration image.

2. The SAR image registration method of claim 1, wherein, Before the step of calculating the overlap rate of the current SAR image and the historical SAR image according to the preset rule, the method further comprises the following steps: respectively reading and analyzing file headers of the current SAR image and the historical SAR image to obtain a first ground sample distance of the current SAR image and a second ground sample distance of the historical SAR image; determining whether the first ground sample distance and the second ground sample distance are consistent, if not, determining a smaller value of the first ground sample distance and the second ground sample distance, and performing down-sampling processing on the SAR image corresponding to the smaller value so that the first ground sample distance and the second ground sample distance are consistent.

3. The SAR image registration method of claim 1, wherein, The preset rule comprises: The first geographical range of the current SAR image is obtained, and the second geographical range of the historical SAR image is obtained; The current SAR image and the historical SAR image are image-stitched based on the first geographical range and the second geographical range to obtain a stitched image, and first geographical coordinates and second geographical coordinates of the current SAR image and the historical SAR image in the stitched image are determined; The overlap rate of the current SAR image and the historical SAR image is calculated based on the first geographical coordinates and the second geographical coordinates.

4. The SAR image registration method of claim 3, wherein, The first geographical range of the current SAR image is obtained, and the second geographical range of the historical SAR image is obtained, comprising: The upper-left latitude and longitude coordinates, the pixel scale, and the image width and height of the current SAR image and the historical SAR image are obtained respectively, and the right-bottom latitude and longitude coordinates are determined based on the upper-left latitude and longitude coordinates, the pixel scale, and the image width and height; The first geographical range of the current SAR image and the second geographical range of the historical SAR image are determined based on the upper-left latitude and longitude coordinates and the right-bottom latitude and longitude coordinates.

5. The SAR image registration method of claim 1, wherein, The initial registration image corresponding to the current SAR image is obtained by registering the first region image based on the second region image using a correlation matching method, comprising: A rectangular data block with a first preset image size is cut from the second region image based on the center point of the second region image and is taken as a reference image, and a rectangular data block with a second preset image size is cut from the first region image based on the center point of the first region image and is taken as a template image The template image is roamed on the reference image, and the first correlation coefficient of each roaming position is calculated, then the maximum correlation coefficient is screened from the first correlation coefficients of all the roaming positions, and the roaming position corresponding to the maximum correlation coefficient is determined as the best matching position; The offset relationship between the current SAR image and the historical SAR image is determined based on the best matching position, and the current SAR image is image-transformed using the offset relationship to obtain the corresponding initial registration image.

6. The SAR image registration method of claim 5, wherein, Each sub-block in the initial registration image is registered based on each sub-block in the historical SAR image using a correlation matching method to obtain a final registration image, comprising: Each sub-block in the initial registration image corresponding to each sub-block in the historical SAR image is determined, and the best matching point coordinates between the two sub-blocks are calculated; It is judged whether the second correlation coefficient corresponding to the best matching point coordinates is not lower than a third preset threshold value, if not, the best matching point coordinates are taken as the control point coordinates of the corresponding sub-block, if lower, the center point coordinates of the corresponding sub-block are taken as the control point coordinates. A first control point set is obtained based on control point coordinates of all sub-blocks, and a first affine transformation relationship between the initial registration image and the historical SAR image is determined according to the first control point set, and then the initial registration image is subjected to image transformation by using the first affine transformation relationship to obtain a final registration image.

7. A SAR image registration apparatus characterized by comprising: The method comprises the steps of: The SAR image acquisition module is configured to acquire a current SAR image sent by the unmanned aerial vehicle, and match a corresponding historical SAR image from a historical image library based on geographic coordinate information of the current SAR image; The overlap rate calculation module is configured to calculate an overlap rate of the current SAR image and the historical SAR image according to a preset rule, and determine whether the overlap rate is not lower than a first preset threshold value; The region image screening module is configured to, if yes, determine an overlap region, and screen a first region image and a second region image from the current SAR image and the historical SAR image respectively based on the overlap region; The initial registration module is configured to register the first region image by using a correlation matching method based on the second region image to obtain an initial registration image corresponding to the current SAR image, and determine whether a corresponding correlation coefficient is greater than a second preset threshold value; The final registration module is configured to, if yes, perform sub-block processing on the initial registration image and the historical SAR image respectively, and register each sub-block of the initial registration image by using a correlation matching method based on each sub-block of the historical SAR image to obtain a final registration image; If no, the current SAR image is registered by using a SIFT feature matching method based on the historical SAR image to obtain a final registration image; wherein the process of registering the first region image by using the SIFT feature matching method based on the second region image to obtain a final registration image comprises: performing filtering enhancement processing on the first region image and the second region image respectively to obtain corresponding gradient images, acquiring a gradient amplitude graph of the gradient images, and constructing a Harris scale space based on the gradient amplitude graph; performing extreme point detection on the Harris scale space under different scales, calculating a gradient histogram of each extreme point, and determining whether the corresponding extreme point is a key point based on the gradient histogram; if it is a key point, generating a feature descriptor based on the key point, and performing key point matching by using a nearest neighbor distance ratio method based on the feature descriptor to screen a matching point set from the historical SAR image; constructing a second control point set based on the matching point set, and determining a second affine transformation relationship between the current SAR image and the historical SAR image according to the second control point set, and then performing image transformation on the current SAR image by using the second affine transformation relationship to obtain a final registration image.

8. An electronic device, comprising: The computer program is saved in the memory, and the processor is configured to execute the computer program to realize the steps of the SAR image registration method according to any one of claims 1 to 6. The computer program is saved in the memory, and the processor is configured to execute the computer program to realize the steps of the SAR image registration method according to any one of claims 1 to 6. ​ 9. A computer-readable storage medium, characterized in that, A computer program product for storing a computer program; wherein the computer program, when executed by a processor, implements the steps of the SAR image registration method according to any one of claims 1 to 6.

Images