Method and system for improving resolution of natural multi-coverage image of vertical rail scanning remote sensing satellite

Abstract

The invention discloses a vertical rail scanning remote sensing satellite natural multi-coverage image resolution improving method and system, and relates to the field of remote sensing image processing. The method solves the problems that due to the fact that the distance between an existing vertical rail scanning imaging sensor and a ground target is increased, image pixels cover a wider ground range, the actually-measured spatial resolution is reduced, and fine detection and recognition of ships, aircrafts and the ground target cannot be supported. SIFT feature point extraction is carried out, and feature points extracted from different images are matched by using a Euclidean nearest distance matching strategy; taking one registered image as a reference, and constructing an initial high-resolution image through up-sampling; establishing an imaging degradation model from a high-resolution image to a low-resolution observation image; calculating a residual image of the simulation image and the real observation image; and back-projecting the residual error back to the high-resolution image space, and updating high-resolution image estimation.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a method and system for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites. Background Art

[0002] Remote sensing imagery is widely used in fields such as geographic resource surveys, environmental monitoring, disaster warning, and land surveying and mapping. Its rapid, large-scale observation of the Earth's surface is crucial and crucial. High-resolution remote sensing imagery is particularly important for high-precision tasks such as object identification, detailed target monitoring, and intelligent analysis.

[0003] Among these systems, vertical-track scanning remote sensing, a key innovation, offers the unique advantage of capturing wide ground coverage areas with high accuracy. Designed to scan wide areas, vertical-track scanning remote sensing utilizes a rotating mechanism to capture sequential images, which are then stitched together to form a continuous panoramic view of the target area. This imaging method enables the acquisition of large-scale regional information in a single pass, easily covering swaths of thousands of kilometers. This significantly reduces revisit time and enhances the ability to rapidly detect ships, aircraft, and ground targets.

[0004] The vertical-track scanning remote sensing system uses a TDI (time-delayed integration) camera to continuously capture surface images by rotating the satellite payload, forming long strips of remote sensing images. To further expand the observation width, the system design arranges multiple CCD sensor modules with a fixed overlap ratio along the track direction, achieving multi-CCD stitching. Because these CCD imaging areas naturally overlap and spatial resolution decreases with increasing roll angle, at large roll angles, the same area is repeatedly covered by multiple scanning strips, resulting in a so-called "natural multiple coverage" characteristic.

[0005] As the roll angle increases, the distance between the vertical track scanning imaging sensor and the ground target increases, causing the image pixels to cover a wider ground area, significantly reducing the measured spatial resolution and severely losing detailed information, making it impossible to support the precise detection and identification of ships, aircraft, and ground targets. Existing single-frame super-resolution methods are mostly based on single-view interpolation or deep learning reconstruction, which cannot fully utilize the information in the overlapping areas of adjacent strips. They are also limited by imaging distortion and illumination changes, resulting in limited restoration effects. Vertical track scanning imaging naturally forms multiple coverage strips, and multiple frames of images with different orientations and resolutions can be obtained in the same area of ​​the object. However, existing processing pipelines generally regard them as redundant or only use them for alignment, and do not fully tap the potential of multi-angle information fusion. Summary of the Invention

[0006] This invention aims to address the problem of existing vertical-track scanning imaging, where the distance between the sensor and ground targets increases, resulting in image pixels covering a wider ground area. This significantly reduces measured spatial resolution, causes severe loss of detail, and is unable to support the precise detection and identification of ships, aircraft, and ground targets. Existing single-frame super-resolution methods, which are mostly based on single-view interpolation or deep learning reconstruction, fail to fully utilize information from overlapping areas of adjacent strips and are limited by imaging distortion and illumination variations, resulting in limited restoration results.

[0007] To solve the above technical problems, the present invention is achieved through the following technical solutions: The present invention proposes a method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites, the method comprising the following steps: Step 1: Calculate the overlap rate of vertical track scanning remote sensing satellite images, perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%, and name the images with an overlap rate of adjacent strips exceeding 50% as vertical track scanning remote sensing satellite natural multiple coverage images; Step 2: Extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input in step 1, and use the Euclidean closest distance matching strategy to match the feature points extracted from different images; Step 3: Based on the feature point pairs obtained by the Euclidean closest distance matching strategy in step 2, the random sampling consistency algorithm is used to eliminate the mismatched points and determine the optimal homography transformation matrix; Step 4: After image registration is completed based on step 3, one of the registered images is used as a reference to construct an initial high-resolution image by upsampling; Step 5: Establish an imaging degradation model from high-resolution images to low-resolution observation images; Step 6: In each iteration, a high-resolution degraded simulated image is obtained according to the established imaging degradation model, and the residual image between the simulated image and the real observed image is calculated; Step 7: Based on the residual image obtained in step 6, back-project the residual back to the high-resolution image space to update the high-resolution image estimate.

[0008] Furthermore, a preferred embodiment is provided, in which the optimal homography transformation matrix determined in step 3 is:

[0009] Among them and Represent the coordinates of the matching feature point pairs in image 1 and image 2 respectively.

[0010] Furthermore, a preferred embodiment is provided, in which the method for constructing the initial high-resolution image in step 4 is:

[0011] in Represents the interpolation upsampling operator.

[0012] Furthermore, a preferred embodiment is provided, in which the method for establishing the imaging degradation model from the high-resolution image to the low-resolution observation image in step 5 is: The high-resolution image is first convolution blurred and then downsampled by proportional bicubic interpolation to obtain a low-resolution image; For the Frame low-resolution observation image , which is generated by the high-resolution image Action fuzzy operator and sampling operators The result after adding noise :

[0013] in represents the convolution operation, is the point spread function of the imaging system, Indicates the desired magnification number Operator for bicubic downsampling, represents the imaging noise; for the ideal model, Ignore.

[0014] Furthermore, a preferred embodiment is provided, in which the method for calculating the residual image between the simulated image and the real observed image in step 6 is:

[0015] in For the Frame original low-resolution image at position The gray value at It is the value of the corresponding location of the simulated low-resolution image generated based on the current high-score estimate.

[0016] Furthermore, a preferred embodiment is provided, in step 7, based on the residual image obtained in step 6, these residuals are back-projected back to the high-resolution image space, and the method for updating the high-resolution image estimate is:

[0017] in is the number of frames of overlapping images, Represents the pixel position in the high-resolution image coordinate system; Indicates that the Error map after upsampling and reprojecting the frame residual image to a high-resolution coordinate system.

[0018] Furthermore, a preferred embodiment is provided, in which the resolution of the image after the high-resolution image estimation is updated in step 7 is increased to twice the resolution of the original image.

[0019] Solution 2: A vertical-track scanning remote sensing satellite natural multi-coverage image resolution enhancement system, the system comprising: The overlap rate calculation module is used to calculate the overlap rate of vertical track scanning remote sensing satellite images, and perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%. Images with an overlap rate of adjacent strips exceeding 50% are named vertical track scanning remote sensing satellite natural multiple coverage images; The feature point matching module is used to extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input by the overlap rate calculation module, and match the feature points extracted from different images using the Euclidean nearest distance matching strategy; The homography transformation matrix confirmation module is used to eliminate the mismatched points based on the feature point pairs obtained by the closest distance matching strategy of the feature point matching module using the random sampling consistency algorithm to determine the optimal homography transformation matrix; The high-resolution image construction module is used to construct the initial high-resolution image by upsampling after the image registration of the homography transformation matrix confirmation module is completed, using one of the registered images as a reference; Imaging degradation model building module, used to build imaging degradation model from high-resolution image to low-resolution observation image; The residual image calculation module is used to obtain a high-resolution degraded simulated image according to the established imaging degradation model in each iteration process, and calculate the residual image between the simulated image and the real observed image; The high-resolution image estimation module is used to back-project the residual images obtained by the residual image calculation module back into the high-resolution image space and update the high-resolution image estimation.

[0020] Solution 3: A computer device includes a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the method described in Solution 1.

[0021] Solution 4: A computer-readable storage medium storing a computer program, wherein the computer program implements the steps of the method described in Solution 1 when executed by a processor.

[0022] The present invention is beneficial in that: The present invention proposes a method and system for improving the resolution of remote sensing images by utilizing multiple coverage information. By fusing image data from different coverage perspectives, resolution enhancement is achieved, thereby improving the system's ability to perceive and recognize ground objects.

[0023] Under large side swing angle conditions, the present invention utilizes the natural multiple coverage advantages of vertical track scanning remote sensing satellites, and realizes super-resolution reconstruction of the area by aligning and fusing multi-frame image information, restoring high-precision ground structure and texture features, thereby enhancing the practical application value of remote sensing images.

[0024] The method for improving the resolution of natural multiple coverage images of vertical-track scanning remote sensing satellites described in the present invention does not rely on additional hardware investment, and the fusion algorithm is efficient and robust.

[0025] The present invention is also applicable to scenarios where vertical track scanning with large side swing angle imaging leads to deterioration of image quality, and can improve the application capability of the entire remote sensing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flowchart of a method for improving the resolution of natural multiple coverage images of a vertical-track scanning remote sensing satellite as described in Implementation Method 1.

[0027] Figure 2 This is a multi-feature grayscale schematic diagram described in the eleventh embodiment.

[0028] Figure 3 Schematic diagram of image super-resolution results with different numbers of iterations as described in the eleventh embodiment.

[0029] Among them, (a) is a schematic diagram of the results of 5 iterations; (b) is a schematic diagram of the results of 10 iterations; (c) is a schematic diagram of the results of 10 iterations.

[0030] Figure 4 This is a schematic diagram of the ground pixel resolution described in the eleventh embodiment.

[0031] Figure 5 This is a schematic diagram of obtaining multiple frames of super-resolved input images by observing the same area at different angles in a simulation environment in the eleventh embodiment.

[0032] Among them, (a) is a schematic diagram of LR1, with a vertical track angle of 65.56° and an along-track angle of 0.00°; (b) is a schematic diagram of LR2, with a vertical track angle of 65.56° and an along-track angle of 9.46°.

[0033] Figure 6 This is a schematic diagram of feature point extraction and matching results in implementation mode eleven.

[0034] Figure 7Schematic diagram of resampling the LR2 image in the reference coordinate system in the eleventh embodiment.

[0035] Figure 8 Schematic diagram of the initial high-resolution image in the eleventh embodiment.

[0036] Wherein, (a) is a schematic diagram of the initial high-resolution image HR1, and (b) is a schematic diagram of the initial high-resolution image HR2.

[0037] Figure 9 This is a schematic diagram of the final super-resolution result in implementation mode eleven.

[0038] Among them, (a) is a schematic diagram of the LR super-resolution result, and (b) is a schematic diagram of the SR super-resolution result. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the implementation methods of this application clearer, the technical solutions in the implementation methods of this application will be clearly and completely described below in combination with the drawings in the implementation methods of this application. Obviously, the described implementation methods are only part of the implementation methods of this application, not all of the implementation methods.

[0040] Implementation method 1: This implementation method proposes a method for improving the resolution of a natural multi-coverage image of a vertical-track scanning remote sensing satellite, the method comprising the following steps: Step 1: Calculate the overlap rate of vertical track scanning remote sensing satellite images, perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%, and name the images with an overlap rate of adjacent strips exceeding 50% as vertical track scanning remote sensing satellite natural multiple coverage images; Step 2: Extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input in step 1, and use the Euclidean closest distance matching strategy to match the feature points extracted from different images; Step 3: Based on the feature point pairs obtained by the Euclidean closest distance matching strategy in step 2, the random sampling consistency algorithm is used to eliminate the mismatched points and determine the optimal homography transformation matrix; Step 4: After image registration is completed based on step 3, one of the registered images is used as a reference to construct an initial high-resolution image by upsampling; Step 5: Establish an imaging degradation model from high-resolution images to low-resolution observation images; Step 6: In each iteration, a high-resolution degraded simulated image is obtained according to the established imaging degradation model, and the residual image between the simulated image and the real observed image is calculated; Step 7: Based on the residual image obtained in step 6, back-project the residual back to the high-resolution image space to update the high-resolution image estimate.

[0041] Implementation 2: This implementation further limits the method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites described in Implementation 1. The optimal homography transformation matrix determined in step 3 is:

[0042] in and Represent the coordinates of the matching feature point pairs in image 1 and image 2 respectively.

[0043] Implementation method 3: This implementation method further limits the method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites described in implementation method 1. The method for constructing the initial high-resolution image in step 4 is:

[0044] in Represents the interpolation upsampling operator.

[0045] Implementation 4: This implementation further defines the method for improving the resolution of natural multi-coverage images from vertical-track scanning remote sensing satellites described in Implementation 1. The method for establishing an imaging degradation model from high-resolution images to low-resolution observation images in step 5 is as follows: The high-resolution image is first convolution blurred and then downsampled by proportional bicubic interpolation to obtain a low-resolution image; For the Frame low-resolution observation image , which is generated by the high-resolution image Action fuzzy operator and sampling operators The result after adding noise :

[0046] in represents the convolution operation, is the point spread function of the imaging system, Indicates the required magnification Operator for bicubic downsampling, represents the imaging noise; for the ideal model, Ignore.

[0047] Implementation 5: This implementation further limits the method for improving the resolution of natural multi-coverage images of vertical-track scanning remote sensing satellites described in Implementation 1. The method for calculating the residual image between the simulated image and the real observed image in step 6 is:

[0048] in For the Frame original low-resolution image at position The gray value at It is the value of the corresponding location of the simulated low-resolution image generated based on the current high-score estimate.

[0049] Implementation 6: This implementation further limits the method for improving the resolution of natural multi-coverage images from vertical-track scanning remote sensing satellites described in Implementation 1. In step 7, based on the residual image obtained in step 6, the method for back-projecting these residuals back into the high-resolution image space and updating the high-resolution image estimate is as follows:

[0050] in is the number of frames of overlapping images, Represents the pixel position in the high-resolution image coordinate system; Indicates that the Error map after upsampling and reprojecting the frame residual image to a high-resolution coordinate system.

[0051] Implementation method seven: This implementation method further limits the method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites described in implementation method one. In step 7, the resolution of the image after updating the high-resolution image estimate is improved to twice the resolution of the original image.

[0052] Embodiment 8: This embodiment proposes a vertical-track scanning remote sensing satellite natural multi-coverage image resolution enhancement system, the system comprising: The overlap rate calculation module is used to calculate the overlap rate of vertical track scanning remote sensing satellite images, and perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%. Images with an overlap rate of adjacent strips exceeding 50% are named vertical track scanning remote sensing satellite natural multiple coverage images; The feature point matching module is used to extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input by the overlap rate calculation module, and match the feature points extracted from different images using the Euclidean nearest distance matching strategy; The homography transformation matrix confirmation module is used to eliminate the mismatched points based on the feature point pairs obtained by the closest distance matching strategy of the feature point matching module using the random sampling consistency algorithm to determine the optimal homography transformation matrix; The high-resolution image construction module is used to construct the initial high-resolution image by upsampling after the image registration of the homography transformation matrix confirmation module is completed, using one of the registered images as a reference; Imaging degradation model building module, used to build imaging degradation model from high-resolution image to low-resolution observation image; The residual image calculation module is used to obtain a high-resolution degraded simulated image according to the established imaging degradation model in each iteration process, and calculate the residual image between the simulated image and the real observed image; The high-resolution image estimation module is used to back-project the residual images obtained by the residual image calculation module back into the high-resolution image space and update the high-resolution image estimation.

[0053] Implementation method 9. This implementation method proposes a computer device, including a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the methods described in implementation methods 1 to 7.

[0054] Embodiment 10: This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the steps of the methods described in Embodiments 1 to 7 are implemented. Implementation 11: This implementation provides an example, which is used to explain the above implementations 1 to 8. Specifically, the example is as follows: See also Figure 1 and Figure 9 To illustrate this embodiment, the embodiment includes the following steps: Step 1: Calculate the overlap rate In order to meet the basic requirement of subsequent multi-frame of vertical-track scanning remote sensing satellite images, that is, the same area is observed at least twice, the overlap rate of vertical-track scanning remote sensing satellite images is calculated, and subsequent super-resolution is performed on images with an overlap rate of adjacent strips exceeding 50%.

[0055] The camera has a field of view of 20° along the track and a sub-satellite resolution of =1m, corresponding to the width of the sub-satellite point along the track direction =176.33km. Based on the ground speed of 7.062km / s and the imaging period of 22.5s, the strip spacing l=158.90km and the overlap rate .

[0056] If the sweep angle of any strip is required When it is greater than a certain value, it is completely covered by the two strips in front and behind. The overlap rate of adjacent strips should be at least 50%. If the width along the track is W, then , we get W>317.8km, and the ground pixel resolution along the track is .

[0057] The ground pixel resolution along the track is proportional to the object distance L. The geometric relationship between the camera imaging center C and the object point D is shown in the figure, where R is the radius of the earth and H is the orbit height. The geocentric angle between C and D can be calculated by The sine theorem in (1) The distance between C and D (object distance) is: (2) Then the ground pixel resolution along the track is (3) Calculation shows that the sweep angle is greater than 53° at this time.

[0058] Step 2: SIFT feature point extraction and matching. In the simulation environment, observe the same area from different angles to obtain multiple frames of super-resolution input images, such as Figure 5 As shown, the resolution of the images is 1024×1024 pixels.

[0059] First, SIFT feature extraction is performed on two overlapping low-resolution images. The SIFT (Scale-Invariant Feature Transform) algorithm extracts image feature points at different scales and calculates their 128-dimensional descriptors, which are insensitive to rotation and scale changes. Subsequently, a Euclidean nearest neighbor matching strategy is used to match feature points within the feature descriptor sets of the two images, generating several corresponding feature point pairs. The feature point extraction and matching results are shown in the figure.

[0060] The average reprojection error of the image is 0.53 pixels.

[0061] Step 3: Estimation of homography matrix and image registration. Using the feature point pairs after mismatching, calculate the homography matrix To achieve accurate geometric registration between images. The homography matrix is ​​a matrix that describes the plane projection transformation. Matrix, which maps points in one image coordinate system to another image coordinate system: (4) in and Represent the coordinates of the matching feature point pairs in image 1 and image 2 respectively.

[0062] For the feature point matching results obtained, due to the similarity of feature points, there may be mismatching problems. By eliminating mismatches through the random sampling consensus algorithm (RANSAC), the optimal Matrix. The calculated homography matrix This is used to accurately resample one image to the coordinate system of the other, aligning the two images to the same reference coordinates (i.e., completing image registration). After registration, the pixel locations of the corresponding features in the two low-resolution images are aligned.

[0063] According to the homography matrix solved by the feature matching result, the LR1 image is used as the reference coordinate system, and the LR2 image is unified to the reference coordinate system. The image of the LR2 image resampled in the reference coordinate system is as follows: Figure 7 As shown: Step 4: Interpolation construction of the initial high-resolution image. After the image registration is completed, one of the registered images is used as a reference to construct the initial high-resolution image through bicubic interpolation upsampling. .remember is the reference image, the initial high-resolution image can be expressed as: (5) in represents the interpolation upsampling operator. This initial high-resolution image provides the starting point for iterative optimization.

[0064] The initial high-resolution images of LR1 and LR2 are as follows Figure 8 As shown: Step 5: Establishment of image degradation model. In order to fuse multi-frame image information to improve resolution, an imaging degradation model from high-resolution image to low-resolution observation image is established. In the remote sensing imaging process, the main factors affecting image clarity include blurring caused by the point spread function of the optical system and downsampling caused by the sampling of the imaging sensor. These processes can be modeled as first performing convolution blur on the high-resolution image, and then downsampling by proportional bicubic interpolation to obtain a low-resolution image. Frame low-resolution observation image , which can be expressed as a high-resolution image Action fuzzy operator and sampling operators The result after adding noise : (6) in represents the convolution operation, is the point spread function of the imaging system, Indicates the required magnification Operator for bicubic downsampling, represents the imaging noise. For the ideal model, The above degradation model describes how a high-resolution image is blurred and downsampled to produce a low-resolution image that is consistent with the actual observation.

[0065] Step 6: Residual image calculation. In the iterative back-projection algorithm, the high-resolution image is continuously adjusted to make the simulated low-resolution image generated by the degradation model consistent with the actual observation. In the iteration, the current high-resolution estimate By projecting the model into each low-resolution image space, the low-resolution image is simulated Then, the simulated image is compared with the real observed image and the difference is calculated to obtain the residual image. Frame image in iteration The residual when is defined as: (7) in For the Frame original low-resolution image at position The gray value at The residual is the value of the simulated low-resolution image at the corresponding position generated based on the current high-score estimate. It reflects the error between the low-score image generated by the imaging model using the current high-score estimation and the actual observation.

[0066] Step 7: High-resolution image update. After calculating the residuals of each frame, these residuals are back-projected back to the high-resolution image space to update the high-resolution image estimate. This step is achieved by upsampling the low-resolution residuals and accumulating them into the high-resolution image, which is equivalent to allocating the observation error back to the high-resolution pixels for correction. Represents The corresponding upsampling operator, for the case of double coverage (two frames of image), averages the upsampling residuals of the two frames and adds them back to the high-resolution image. The update formula is as follows: (8) in The number of frames of overlapping images (when double overlay ), Represents the pixel position in the high-resolution image coordinate system. Indicates that the Error map after upsampling and reprojecting the frame residual image to the high-resolution coordinate system. Formula (8) shows that the update amount of the high-resolution image comes from the average result of the back-projected residuals of each low-resolution image. By backpropagating the residuals to the high-resolution grid and accumulating them to the current estimate, the above iterative process gradually corrects the high-resolution image, making the projected simulated low-resolution image closer to the actual observation.

[0067] In summary, see Figure 9As shown in FIG, after the vertical track scanning remote sensing satellite natural multiple coverage image resolution enhancement method and system, the resolution of the super-resolution image is increased to twice the original image resolution.

[0068] Those skilled in the art will understand that the above description is only a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims of the present disclosure may be combined or coupled in various ways, even if such a combination or coupling is not explicitly described in the present disclosure. It is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

[0069] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A method for improving the resolution of natural multi-coverage images from vertical-track scanning remote sensing satellites, characterized in that: The method comprises the following steps: Step 1: Calculate the overlap rate of vertical track scanning remote sensing satellite images, perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%, and name the images with an overlap rate of adjacent strips exceeding 50% as vertical track scanning remote sensing satellite natural multiple coverage images; Step 2: Extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input in step 1, and use the Euclidean closest distance matching strategy to match the feature points extracted from different images; Step 3: Based on the feature point pairs obtained by the Euclidean closest distance matching strategy in step 2, the random sampling consistency algorithm is used to eliminate the mismatched points and determine the optimal homography transformation matrix; Step 4: After image registration is completed based on step 3, one of the registered images is used as a reference to construct an initial high-resolution image by upsampling; Step 5: Establish an imaging degradation model from high-resolution images to low-resolution observation images; Step 6: In each iteration, a high-resolution degraded simulated image is obtained according to the established imaging degradation model, and the residual image between the simulated image and the real observed image is calculated; Step 7: Based on the residual image obtained in step 6, back-project the residual back to the high-resolution image space to update the high-resolution image estimate.

2. The method for improving the resolution of natural multiple coverage images of vertical orbit scanning remote sensing satellites according to claim 1, characterized in that: The optimal homography transformation matrix determined in step 3 is: in and Represent the coordinates of the matching feature point pairs in image 1 and image 2 respectively.

3. The method for improving the resolution of natural multiple coverage images of vertical orbit scanning remote sensing satellites according to claim 1, characterized in that: The method for constructing the initial high-resolution image in step 4 is: in Represents the interpolation upsampling operator.

4. The method for improving the resolution of natural multiple coverage images of vertical orbit scanning remote sensing satellites according to claim 1, characterized in that: The method for establishing the imaging degradation model from high-resolution images to low-resolution observation images in step 5 is: The high-resolution image is first convolution blurred and then downsampled by proportional bicubic interpolation to obtain a low-resolution image; For the Frame low-resolution observation image , which is generated by the high-resolution image Action fuzzy operator and sampling operators The result after adding noise : in represents the convolution operation, is the point spread function of the imaging system, Indicates the required magnification Double three The downsampling operator, represents the imaging noise; for the ideal model, Ignore.

5. The method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites according to claim 1, characterized in that: The method for calculating the residual image between the simulated image and the real observed image in step 6 is: in For the Frame original low-resolution image at position The gray value at It is the value of the corresponding location of the simulated low-resolution image generated based on the current high-score estimate.

6. The method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites according to claim 1, characterized in that: In step 7, based on the residual image obtained in step 6, these residuals are back-projected back to the high-resolution image space, and the method for updating the high-resolution image estimate is: in is the number of frames of overlapping images, Represents the pixel position in the high-resolution image coordinate system; Indicates that the Error map after upsampling and reprojecting the frame residual image to a high-resolution coordinate system.

7. The method for improving the resolution of natural multiple coverage images of vertical track scanning remote sensing satellites according to claim 1, characterized in that: After updating the high-resolution image estimate in step 7, the resolution of the image is increased to twice the original image resolution.

8. A vertical-track scanning remote sensing satellite natural multi-coverage image resolution enhancement system, characterized by: The system comprises: The overlap rate calculation module is used to calculate the overlap rate of vertical track scanning remote sensing satellite images, and perform super-resolution on images with an overlap rate of adjacent strips exceeding 50%. Images with an overlap rate of adjacent strips exceeding 50% are named vertical track scanning remote sensing satellite natural multiple coverage images; The feature point matching module is used to extract SIFT feature points from the vertical track scanning remote sensing satellite natural multiple coverage image input by the overlap rate calculation module, and match the feature points extracted from different images using the Euclidean nearest distance matching strategy; The homography transformation matrix confirmation module is used to eliminate the mismatched points based on the feature point pairs obtained by the closest distance matching strategy of the feature point matching module using the random sampling consistency algorithm to determine the optimal homography transformation matrix; The high-resolution image construction module is used to construct the initial high-resolution image by upsampling after the image registration of the homography transformation matrix confirmation module is completed, using one of the registered images as a reference; Imaging degradation model building module, used to build imaging degradation model from high-resolution image to low-resolution observation image; The residual image calculation module is used to obtain a high-resolution degraded simulated image according to the established imaging degradation model in each iteration process, and calculate the residual image between the simulated image and the real observed image; The high-resolution image estimation module is used to back-project the residual images obtained by the residual image calculation module back into the high-resolution image space and update the high-resolution image estimation.

9. A computer device comprising a memory and a processor, characterized in that A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images