Full-automatic remote sensing image geometric fine correction method and equipment based on image registration

Abstract

The invention discloses a full-automatic remote sensing image geometric fine correction method and equipment based on image registration, and relates to the technical field of remote sensing image processing. The image registration-based full-automatic geometric fine correction method for the remote sensing image mainly comprises the following steps of: obtaining a boundary vector of a target remote sensing image by utilizing the target remote sensing image to be corrected and coordinates of four corners of the target remote sensing image; and obtaining a reference image according to the four-corner point coordinates and the boundary vector, and correcting the to-be-corrected target remote sensing image by using a local geometric fine correction method to obtain a corrected image. According to the full-automatic remote sensing image geometric fine correction method and equipment based on image registration provided by the invention, geometric fine correction of the remote sensing image with a sub-pixel-level precision result can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and more specifically to a fully automatic remote sensing image geometric precision correction method and equipment based on image registration. Background Art

[0002] After the original remote sensing images have been preprocessed by radiation correction, geometric correction, atmospheric correction, etc., the remote sensing images obtained often still have certain deviations from the actual ground coordinates; and due to the error in accuracy calculation, in the same geographic coordinate system, remote sensing images with geometric accuracy obtained by different sensors will also have certain geographic location deviations. Geometric precision correction, also known as geometric registration, refers to using images, maps or data sets with geometric accuracy as references, using the ground control point data in these references to mathematically simulate the geometric deviations of remote sensing images, and establishing a mathematical model of the coordinate relationship between the corresponding points of the image before and after correction to correct the remote sensing image. Accurate geometric precision correction is a basic requirement for many remote sensing applications, such as quantitative analysis of remote sensing images, change detection, image fusion, map measurement or update, etc.

[0003] One of the key difficulties of geometric precision correction technology is the selection of control points in the original remote sensing image and the remote sensing image with geometric accuracy, which directly determines the effect and quality of the correction result. At present, there are two main control point selection methods: feature-based control point selection and intensity-based control point selection. The feature-based control point selection method generally uses the positions of ground objects with different identification features as control points, such as intersections, buildings, boundaries of farmland, or other objects with clear contours that are not easy to change; the intensity-based control point selection method relies on the recognition of similar grayscale value patterns that appear in the reference data set and the target remote sensing image. The feature-based control point selection method usually has a low computational cost; but each different remote sensing image requires a certain number of control points with identification features, and their annotation takes a lot of time. In addition, when the identification features in the remote sensing image are unevenly distributed, the calculated correction model may not be applicable to the entire image. In contrast, although the intensity-based control point selection method has a high computational cost, it does not rely on the existence of image recognition features and can achieve sub-pixel accuracy under the premise that the image is not overly distorted.

[0004] In 2017, Daniel Scheffler et al. proposed a geometric precision correction technology based on intensity-selected control points - the AROSICS (Automatic and Robust Open-Source Image Co-Registration Software) method. This method uses phase correlation technology to detect sub-pixel displacement to automatically find control points in the image, and then filters out control points that do not meet the requirements through validity checks, threshold checks, reliability checks, similarity checks, and outlier checks. The geometric offset is calculated for the qualified control points, and the affine transformation model is fitted to perform geometric precision correction on the target remote sensing image. The joint geometric registration results of the AROSICS method on multiple sensors can reach the sub-pixel level, and have been used in many scientific studies for the geometric registration of remote sensing images such as Sentinel-2, Landsat-5, and Landsat-7.

[0005] The AROSICS method can basically realize the automation of control point selection, geometric offset calculation and geometric offset correction, but it requires a certain amount of labor cost in the data preparation part. For a target remote sensing image, before using the AROSICS method for geometric precision correction, it is necessary to obtain a remote sensing image with a geographical range not less than the image to be corrected, a low degree of distortion and geometric accuracy as a reference image, and the screening of reference images usually takes a lot of time; and with the rapid growth of the amount of available remote sensing data in recent years and the gradual increase in the demand for further remote sensing image geometric registration, it is very necessary to realize the full automation of remote sensing image geometric precision correction in practical applications. Summary of the invention

[0006] The purpose of the present invention is to provide a fully automatic remote sensing image geometric precision correction method and equipment based on image registration, which can achieve geometric precision correction of remote sensing images with sub-pixel accuracy results.

[0007] The present invention provides a fully automatic remote sensing image geometric precision correction method based on image registration, comprising the following steps: S1: acquiring a target remote sensing image to be corrected and the coordinates of its four corner points, and obtaining a boundary vector of the target remote sensing image according to the coordinates of the four corner points; S2: obtaining a reference image according to the coordinates of the four corner points and the boundary vector; S3: correcting the target remote sensing image to be corrected by using a local geometric precision correction method according to the reference image to obtain a corrected image.

[0008] Furthermore, step S2 specifically includes: selecting a reference image product from Google Earth Engine according to the boundary vector; determining a download date range, projection, band and resolution of the reference image according to the reference image product; and remotely downloading the reference image from Google Earth Engine using an image download function of a geemap library according to the download date range, projection, band and resolution to obtain a reference image.

[0009] Furthermore, step S2 can also be implemented in the following manner: obtain a reference image database, filter the reference image database according to the four corner point coordinates, and obtain a hit image; if the number of hit images is greater than 1, merge the hit images according to the four corner point coordinates to obtain a merged hit image; if the number of hit images is 1, the hit image is a merged hit image; according to the boundary vector, crop the merged hit image to obtain a reference image.

[0010] Furthermore, step S3 specifically includes: S31: determining the local matching window size, and obtaining the target remote sensing image subset image and the reference image subset image according to the local matching window size, the target remote sensing image to be corrected, and the reference image; S32: obtaining a control point grid according to the target remote sensing image subset image and the reference image subset image; S33: verifying and screening the control points in the control point grid to obtain a final control point grid; S34: using the final control point grid to perform offset correction on the target remote sensing image to be corrected to obtain a corrected image.

[0011] Further, step S32 specifically includes: S321: according to the target remote sensing image subset image and the reference image subset image, using the FFTW method, obtaining an integer offset; S322: according to the integer offset, adjusting the angular coordinates of the matching window to obtain a new position, and according to the new position, recreating the subset image of the target image to obtain a new target remote sensing image subset image; S323: verifying the validity of the integer offset to obtain a new integer offset, and using the new integer offset to correct the new target remote sensing image subset image to obtain a target remote sensing image subset image with the integer offset corrected; S324: according to the target remote sensing image subset image with the integer offset corrected and the reference image subset image, obtaining a sub-pixel offset, and according to the sub-pixel offset, obtaining a control point grid.

[0012] Further, step S321 specifically includes: S3211: using the FFTW method, transforming the target remote sensing image subset image and the reference image subset image into the frequency domain to obtain the target remote sensing image subset image spectrum and the reference image subset image spectrum; S3212: obtaining the cross-power spectrum based on the obtained target remote sensing image subset image spectrum and the reference image subset image spectrum; S3213: using the inverse FFTW method to transform the cross-power spectrum into the spatial domain to obtain the spatial domain cross-power spectrum; S3214: normalizing the obtained spatial domain cross-power spectrum to obtain the spatial domain normalized cross-power spectrum; S3215: based on the spatial domain normalized cross-power spectrum, taking the target remote sensing image subset image as a reference, calculating the distance between the peak position and the center of the spatial domain normalized cross-power spectrum in the horizontal and vertical directions, and obtaining the integer offsets in the horizontal and vertical directions respectively.

[0013] Further, step S324 specifically includes: S3241: performing frequency domain transformation on the target remote sensing image subset image corrected for integer offset and the reference image subset image to obtain a new cross-power spectrum; S3242: obtaining a sub-pixel offset based on the new cross-power spectrum; S3243: obtaining a total map offset and an absolute map coordinate offset based on the sub-pixel offset and the reference image subset image; obtaining a control point grid based on the total map offset and the absolute map coordinate offset.

[0014] Further, step S33 specifically includes: S331: performing threshold verification on the control points in the control point grid to obtain a first control point grid; S332: performing reliability verification on the control points in the first control point grid to obtain a second control point grid; S333: performing MSSIM verification on the control points in the second control point grid to obtain a third control point grid; S334: performing RANSAC verification on the control points in the third control point grid to obtain a final control point grid.

[0015] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned fully automatic remote sensing image geometric precision correction method based on image registration are implemented.

[0016] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned fully automatic remote sensing image geometric precision correction method based on image registration are implemented.

[0017] The method and device for fully automatic remote sensing image geometric precision correction based on image registration provided by the present invention have the following beneficial effects: The present invention uses a target remote sensing image to be corrected and the coordinates of its four corner points to obtain the boundary vector of the target remote sensing image; obtains a reference image according to the coordinates of the four corner points and the boundary vector, and uses a local geometric precision correction method to correct the target remote sensing image to obtain a corrected image; the present invention can realize the geometric precision correction of remote sensing images with sub-pixel precision results, improve the calculation efficiency, and realize the full automation of the geometric precision correction of remote sensing images, and only needs to input the target remote sensing image and select a reference image product with geometric precision, so as to fully automatically perform geometric precision correction on the target image and output the corrected result of the target remote sensing image; the present invention can be deployed in a server environment or a conventional computer; it can be implemented based on high-level languages ​​such as Python, has a friendly interface, can be simply integrated into the existing remote sensing processing flow, and has strong usability and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 It is a flow chart of a fully automatic remote sensing image geometric precision correction method based on image registration provided by the present invention; Figure 2 It is an algorithm flow chart of the fully automatic remote sensing image geometric precision correction method based on image registration provided by the present invention; Figure 3 is an input image example provided by the present invention; Figure 4 is an output image example provided by the present invention; Figure 5 This is an example diagram of the visualization of the result accuracy of the online remote sensing image geometric precision correction method provided by the present invention; Figure 6 This is an example diagram of the visualization of the accuracy of the results of the offline remote sensing image geometric precision correction method provided by the present invention; Figure 7 It is a structural block diagram of the computer device provided by the present invention. DETAILED DESCRIPTION

[0019] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0020] Figure 1 A schematic diagram of a fully automatic remote sensing image geometric precision correction method based on image registration in this embodiment is shown. In this embodiment, the fully automatic remote sensing image geometric precision correction method based on image registration includes the following steps: S1: Obtain the target remote sensing image to be corrected and the coordinates of its four corner points, and obtain the boundary vector of the target remote sensing image according to the coordinates of the four corner points; As an embodiment, in step S1, a target remote sensing image to be corrected is input, coordinates of four corner points of the target remote sensing image are obtained, and a boundary vector of the target remote sensing image is generated; S2: Obtain a reference image according to the coordinates of the four corner points and the boundary vector; In one embodiment, step S2 specifically includes: selecting a reference image product from Google Earth Engine according to the boundary vector; determining a download date range, projection, band, and resolution of the reference image according to the reference image product; remotely downloading the reference image from Google Earth Engine using an image download function of a geemap library according to the download date range, projection, band, and resolution to obtain a reference image; In another embodiment, step S2 specifically includes: obtaining a reference image database, screening the reference image database according to the four corner point coordinates to obtain a hit image; if the number of hit images is greater than 1, merging the hit images according to the four corner point coordinates to obtain a merged hit image; if the number of hit images is 1, the hit image is a merged hit image; according to the boundary vector, cropping the merged hit image to obtain a reference image; It should be noted that the fully automatic remote sensing image geometric precision correction method based on image registration proposed in the present invention can be used for online remote sensing image geometric precision correction or offline remote sensing image geometric precision correction. The difference between the two is whether to download the reference image for geometric precision correction online. When obtaining the reference image, the online remote sensing image geometric precision correction method remotely downloads the reference image from Google Earth Engine based on the geemap library, while the offline remote sensing image geometric precision correction method cuts out a reference image whose range is not less than the target remote sensing image from the existing remote sensing image data set stored locally. S3: According to the reference image, the target remote sensing image to be corrected is corrected by using a local geometric precision correction method to obtain a corrected image; As an embodiment, step S3 specifically includes: S31: determining a local matching window size, and obtaining a target remote sensing image subset image and a reference image subset image according to the local matching window size, the target remote sensing image to be corrected, and the reference image; S32: obtaining a control point grid according to the target remote sensing image subset image and the reference image subset image; As an embodiment, step S32 specifically includes: S321: Obtaining an integer offset using the FFTW method according to the target remote sensing image subset image and the reference image subset image; As an embodiment, step S321 specifically includes: S3211: using the FFTW method, transforming the target remote sensing image subset image and the reference image subset image into the frequency domain to obtain the target remote sensing image subset image spectrum and the reference image subset image spectrum; S3212: obtaining the cross power spectrum according to the obtained target remote sensing image subset image spectrum and the reference image subset image spectrum; S3213: using the inverse FFTW method to perform spatial domain transformation on the cross power spectrum to obtain the spatial domain cross power spectrum; S3214: normalizing the obtained spatial domain cross power spectrum to obtain the spatial domain normalized cross power spectrum; S3215: according to the spatial domain normalized cross power spectrum, taking the target remote sensing image subset image as a reference, calculating the distance between the peak position and its center in the spatial domain normalized cross power spectrum in the horizontal and vertical directions, and obtaining the integer offsets in the horizontal and vertical directions respectively; Specifically, in step S321, the integer offset in the matching window is calculated, which specifically includes: using the world's fastest Fourier Transform (Fastest Fourier Transform in the West, FFTW) to transform the target image and the subset image of the reference image into the frequency domain; generating a cross power spectrum based on the phase correlation of the two subset images in the frequency domain; using inverse FFTW to transform the cross power spectrum back to the spatial domain; normalizing the cross power spectrum to obtain a normalized cross power spectrum in the spatial domain; in the spatial domain, taking the subset image of the target image as a reference, calculating the distance between the obvious peak position in the normalized cross power spectrum in the X and Y directions and its center, and obtaining the integer offsets in the X and Y directions respectively; S322: adjusting the angular coordinates of the matching window according to the integer offset to obtain a new position, and recreating the subset image of the target image according to the new position to obtain a new subset image of the target remote sensing image; Specifically, in step S322, the angular coordinates of the matching window are adjusted according to the calculated integer offset and the subset image of the target image is recreated at the new position; S323: verifying the validity of the integer offset to obtain a new integer offset, and using the new integer offset to correct the new target remote sensing image subset image to obtain the target remote sensing image subset image corrected with the integer offset; Specifically, in step S323, the calculated integer offset is verified for validity, including: (1) setting the initial value of the number of iterations to 0; (2) calculating the integer offset of the subset image of the target image, and determining whether the integer offset is 0 at this time, and if so, going to step S324; if not, going to (3); (3) determining whether the number of iterations is greater than 5, and if so, ending the iteration, and if not, going to (4); (4) correcting the subset image of the target image according to the new integer offset; (5) increasing the number of iterations by 1, and repeating (2) to (5); S324: obtaining a sub-pixel offset according to the target remote sensing image subset image corrected for integer offset and the reference image subset image, and obtaining a control point grid according to the sub-pixel offset; As an embodiment, step S324 specifically includes: S3241: performing frequency domain transformation on the target remote sensing image subset image corrected for integer offset and the reference image subset image to obtain a new cross power spectrum; S3242: Obtain a sub-pixel offset according to the new cross-power spectrum; Specifically, step S3242 includes: obtaining a sub-pixel offset according to the new cross power spectrum, such as the formula: , in, and are the sub-pixel offsets in the horizontal and vertical directions, represents the peak position of the new cross-power spectrum, and Respectively represent the adjacent pixel positions in the abscissa and ordinate directions; S3243: Obtaining a total map offset and an absolute map coordinate offset according to the sub-pixel offset and the reference image subset image; and obtaining a control point grid according to the total map offset and the absolute map coordinate offset; Specifically, in step S3243, the calculated sub-pixel offsets are added, and the total offset of the pixel unit relative to the ground sampling distance of the reference image is converted into a map unit to obtain a total map offset and an absolute map coordinate offset; and a control point grid is generated by combining the total map offset with the absolute map coordinate offset; S33: verifying and screening the control points in the control point grid to obtain a final control point grid; As an embodiment, step S33 specifically includes: S331: performing threshold verification on the control points in the control point grid to obtain a first control point grid; Specifically, in step S331, the offset of more than 5 reference image pixels is eliminated; S332: Perform reliability verification on the control points in the first control point grid to obtain a second control point grid; As an embodiment, step S332 specifically includes: S3321: According to the first control point grid, obtain the reliability percentage of each control point, such as the formula: , , , , in, represents the control point, represents the control point in the cross power spectrum The average value of the peak values ​​in the surrounding unit pixels, represents the number of pixels in the cross-power spectrum, and Represents control points The horizontal and vertical coordinates of represents the control point in the cross power spectrum The average value of the peak values ​​outside the surrounding unit pixels, represents the control point in the cross power spectrum The standard deviation of the peak value outside the surrounding unit pixels, represents the reliability percentage; Specifically, in step S3321, the sharpness of the three-dimensional shape quantization peak of the cross power spectrum is analyzed to evaluate the measure of the reliability of each control point; S3322: If the reliability percentage of the control point is not less than a preset reliability percentage threshold, retain the control point; otherwise, delete the control point to obtain a second control point grid; As an embodiment, the preset reliability percentage threshold is 30; S333: Perform MSSIM verification on the control points in the second control point grid to obtain a third control point grid; Specifically, in step S333, the mean structural similarity index (MSSIM) is used to calculate the similarity of the subset image content in the matching window before and after the offset correction, specifically including: for each control point in the matching window, the MSSIM of the subset image of the target image before and after the sub-pixel offset correction is calculated respectively; it is determined whether the MSSIM value of each control point after the sub-pixel offset correction is reduced, if so, the control point is deleted, if not, the control point is retained; S334: performing RANSAC verification on the control points in the third control point grid to obtain a final control point grid; Specifically, in step S334, a random sampling consensus algorithm (RANSAC) is used to calculate the parameters of the affine transformation between the reference image and the target image, and outliers in the previously calculated offset vector grid are filtered out; S34: Using the final control point grid, offset correction is performed on the target remote sensing image to be corrected to obtain a corrected image.

[0021] Furthermore, the fully automatic remote sensing image geometric precision correction method based on image registration also includes outputting the target remote sensing image correction result; end.

[0022] In some embodiments, the above-mentioned fully automatic remote sensing image geometric precision correction method based on image registration can also be implemented in the following manner. Figure 2 The present invention provides an algorithm flow chart of a fully automatic remote sensing image geometric precision correction method based on image registration.

[0023] In one embodiment, a fully automatic remote sensing image geometric precision correction method based on image registration performs geometric precision correction on online remote sensing images, comprising the following steps: Step 1: Input the target remote sensing image to be corrected. In this embodiment, an Aster image taken in June 2018 is selected, including the B1, B2, and B3 bands. Figure 3 (a) shows the target remote sensing image to be corrected; Step 2: Obtain the coordinates of the four corner points of the target remote sensing image and generate the boundary vector of the target remote sensing image; Step 3: Download the reference image based on the geemap according to the boundary vector. This step specifically includes: Step 3.1: Select a reference image product from Google Earth Engine. In this embodiment, the Collection 2 Tier 1 TOA Reflectance product of Landsat 8 is selected. This product has been geometrically corrected and has a certain geometric accuracy. Step 3.2: Determine the download date range, projection, band and resolution of the reference image according to the target remote sensing image; considering the similarity with the target image, this embodiment limits the acquisition date range of the reference image to 2018-01-01 to 2018-06-01; the projection is the same as the target image, EPSG: 32649; the band is all bands of Landsat 8, B1 ~ B11; the resolution is 30 meters; Step 3.3: Use the image download function of the geemap library to remotely download the reference image from Google Earth Engine. The reference image obtained by downloading is as follows: Figure 3 (b) shows the reference image downloaded remotely from Google Earth Engine via geemap in the online remote sensing image geometric precision correction method; Step 4: Correct the target image based on the local geometric precision correction method of AROSICS. This step specifically includes: Step 4.1: Determine the local matching window size, and divide the target image and the reference image into a subset image set according to the matching window size. This embodiment uses a default matching window size of 256×256; Step 4.2: Calculate the geometric offset of the target remote sensing image and generate a control point grid, obtaining a total of 6907 control points; Step 4.3: Verify and screen the control points, where 1455 control points are screened out by reliability verification, 378 control points are screened out by MSSIM verification, and 502 control points are screened out by RANSAC verification, and finally 4572 control points are retained; Step 4.4: Perform offset correction on the target image based on the control point grid that is finally verified to be available; Step 5: Output the target remote sensing image correction result, such as Figure 4 (a) shows the geometric precision correction result of the target remote sensing image by the online remote sensing image geometric precision correction method; Step 6: Calculate the offset between the control point grids of the target remote sensing image before and after correction and the reference image, and visualize them, as shown in the figure below. Figure 5 The result of the online remote sensing image geometric precision correction method is visualized in meters. Figure 5 (a) is the offset between the target remote sensing image before correction and the reference image. Figure 5 (b) is the offset between the target remote sensing image after correction and the reference image.

[0024] In another embodiment, the fully automatic remote sensing image geometric precision correction method based on image registration is performed by offline remote sensing image geometric precision correction, including the following steps: Step 1: Input the target remote sensing image to be corrected. The target image used in this embodiment is the same as the online remote sensing image geometric precision correction; Step 2: Obtain the coordinates of the four corner points of the target remote sensing image and generate the boundary vector of the target remote sensing image; Step 3: Crop the reference image from the database according to the boundary vector. This step specifically includes: Step 3.1: According to the coordinates of the four corner points of the target remote sensing image, the images hit in the reference data set are screened; the reference data set of the database is the Landsat 8 image of Hubei Province, which is stored according to the coordinates of the four corner points; the target remote sensing image used hits four reference images; Step 3.2: Because the number of reference images hit by screening is greater than 1, this embodiment merges the reference images according to the coordinates to finally obtain a reference image; Step 3.3: Crop the reference image according to the boundary vector of the target image, and finally obtain a reference image with the same boundary range as the target image, such as Figure 3(c) shows the reference image cropped from the reference data set in the offline remote sensing image geometric precision correction method; Step 4: Correct the target image based on the local geometric precision correction method of AROSICS. This step specifically includes: Step 4.1: Determine the local matching window size, and divide the target image and the reference image into a subset image set according to the matching window size. This embodiment uses a default matching window size of 256×256; Step 4.2: Calculate the geometric offset of the target remote sensing image and generate a control point grid, obtaining a total of 6948 control points; Step 4.3: Verify and screen the control points, where 549 control points are screened out by reliability verification, 39 control points are screened out by MSSIM verification, and 671 control points are screened out by RANSAC verification, and finally 5689 control points are retained; Step 4.4: Perform offset correction on the target image based on the last verified available control point grid; Step 5: Output the target remote sensing image correction result, such as Figure 4 (b) shows the geometric precision correction result of the target remote sensing image by the offline remote sensing image geometric precision correction method; Step 6: Calculate the offset between the control point grids of the target remote sensing image before and after correction and the reference image, and visualize them, as shown in the figure below. Figure 6 The result of the offline remote sensing image geometric precision correction method is visualized in meters. Figure 6 (a) is the offset between the target remote sensing image before correction and the reference image. Figure 6 (b) is the offset between the target remote sensing image after correction and the reference image.

[0025] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned fully automatic remote sensing image geometric precision correction method based on image registration are implemented. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0026] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned fully automatic remote sensing image geometric precision correction method based on image registration are implemented.

[0027] like Figure 7 As shown, the computer device may include: at least one processor 121, such as a central processing unit (CPU), at least one communication interface 123, a memory 124, and at least one communication bus 122. Among them, the communication bus 122 is used to realize the connection and communication between these components. Among them, the communication interface 123 may include a display screen (Display), a keyboard (Keyboard), and the optional communication interface 123 may also include a standard wired interface and a wireless interface. The memory 124 may be a high-speed random access memory (Random Access Memory, RAM), or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 124 may also be at least one storage device located away from the aforementioned processor 121. Among them, the memory 124 stores an application program, and the processor 121 calls the program code stored in the memory 124 to perform any of the above method steps. Among them, the communication bus 122 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 122 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7Only one line is used to represent it, but it does not mean that there is only one bus or one type of bus. Among them, the memory 124 may include a volatile memory (volatile memory), such as a random-access memory (random-access memory, RAM); the memory may also include a non-volatile memory (non-volatile memory), such as a flash memory (flash memory), a hard disk drive (hard disk drive, HDD) or a solid-state drive (solid-state drive, SSD); the memory 124 may also include a combination of the above-mentioned types of memory. Among them, the processor 121 may be a central processing unit (central processing unit, CPU), a network processor (network processor, NP) or a combination of CPU and NP. Among them, the processor 121 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit (application-specific integrated circuit, ASIC), a programmable logic device (programmable logic device, PLD) or a combination thereof. The above-mentioned PLD may be a complex programmable logic device (complex programmable logic device, CPLD), a field-programmable gate array (field-programmable gate array, FPGA), a generic array logic (generic array logic, GAL) or any combination thereof. Optionally, the memory 124 is also used to store program instructions. The processor 121 can call the program instructions to implement the fully automatic remote sensing image geometric precision correction method based on image registration as in this embodiment.

[0028] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A fully automatic remote sensing image geometric precision correction method based on image registration, characterized in that: The following steps are involved: S1: Obtain the target remote sensing image to be corrected and the coordinates of its four corner points, and obtain the boundary vector of the target remote sensing image according to the coordinates of the four corner points; S2: Obtain a reference image according to the coordinates of the four corner points and the boundary vector; S3: According to the reference image, the target remote sensing image to be corrected is corrected by using a local geometric precision correction method to obtain a corrected image.

2. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 1 is characterized in that: Step S2 specifically includes: selecting a reference image product from Google Earth Engine according to the boundary vector; determining a download date range, projection, band and resolution of the reference image according to the reference image product; and remotely downloading the reference image from Google Earth Engine using an image download function of a geemap library according to the download date range, projection, band and resolution to obtain a reference image.

3. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 1 is characterized in that: Step S2 specifically includes: obtaining a reference image database, screening the reference image database according to the four corner point coordinates to obtain a hit image; if the number of hit images is greater than 1, merging the hit images according to the four corner point coordinates to obtain a merged hit image; if the number of hit images is 1, the hit image is a merged hit image; according to the boundary vector, cropping the merged hit image to obtain a reference image.

4. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 1 is characterized in that: Step S3 specifically includes: S31: determining a local matching window size, and obtaining a target remote sensing image subset image and a reference image subset image according to the local matching window size, the target remote sensing image to be corrected, and the reference image; S32: obtaining a control point grid according to the target remote sensing image subset image and the reference image subset image; S33: verifying and screening the control points in the control point grid to obtain a final control point grid; S34: Using the final control point grid, offset correction is performed on the target remote sensing image to be corrected to obtain a corrected image.

5. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 4 is characterized in that: Step S32 specifically includes: S321: Obtaining an integer offset using the FFTW method according to the target remote sensing image subset image and the reference image subset image; S322: adjusting the angular coordinates of the matching window according to the integer offset to obtain a new position, and recreating the subset image of the target image according to the new position to obtain a new subset image of the target remote sensing image; S323: verifying the validity of the integer offset to obtain a new integer offset, and using the new integer offset to correct the new target remote sensing image subset image to obtain the target remote sensing image subset image corrected with the integer offset; S324: Obtain sub-pixel offsets according to the target remote sensing image subset image corrected for integer offsets and the reference image subset image, and obtain a control point grid according to the sub-pixel offsets.

6. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 5 is characterized in that: Step S321 specifically includes: S3211: transforming the target remote sensing image subset image and the reference image subset image into the frequency domain by using the FFTW method to obtain the target remote sensing image subset image spectrum and the reference image subset image spectrum; S3212: Obtain a cross power spectrum according to the obtained target remote sensing image subset image spectrum and the reference image subset image spectrum; S3213: Performing spatial domain transformation on the cross power spectrum using an inverse FFTW method to obtain a spatial domain cross power spectrum; S3214: normalizing the obtained spatial domain cross power spectrum to obtain a spatial domain normalized cross power spectrum; S3215: According to the spatial domain normalized cross power spectrum, taking the target remote sensing image subset image as a reference, calculate the distance between the peak position and its center in the spatial domain normalized cross power spectrum in the horizontal and vertical directions, and obtain integer offsets in the horizontal and vertical directions respectively.

7. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 5 is characterized in that: Step S324 specifically includes: S3241: performing frequency domain transformation on the target remote sensing image subset image corrected for integer offset and the reference image subset image to obtain a new cross power spectrum; S3242: Obtain a sub-pixel offset according to the new cross-power spectrum; S3243: Obtain a total map offset and an absolute map coordinate offset according to the sub-pixel offset and the reference image subset image; obtain a control point grid according to the total map offset and the absolute map coordinate offset.

8. The fully automatic remote sensing image geometric precision correction method based on image registration according to claim 4 is characterized in that: Step S33 specifically includes: S331: performing threshold verification on the control points in the control point grid to obtain a first control point grid; S332: Perform reliability verification on the control points in the first control point grid to obtain a second control point grid; S333: Perform MSSIM verification on the control points in the second control point grid to obtain a third control point grid; S334: Perform RANSAC verification on the control points in the third control point grid to obtain a final control point grid.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fully automatic remote sensing image geometric precision correction method based on image registration as described in any one of claims 1 to 8 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the fully automatic remote sensing image geometric precision correction method based on image registration as described in any one of claims 1-8 are implemented.

Images