Vertical rail scanning remote sensing satellite image splicing method and system

Abstract

The invention discloses a vertical rail scanning remote sensing satellite image splicing method and system, and belongs to the technical field of remote sensing image processing. The method is used for solving many challenge problems, including inter-strip positioning deviation, image distortion, scale difference and discontinuous brightness color, faced by an existing remote sensing image splicing method in fusion processing of vertical rail scanning multi-strip images. According to the method, seamless fusion is completed through feature extraction and matching, RANSAC matching optimization, optimal splicing line generation, seam area feathering processing and seamless image output, robust matching and homography transformation are obtained by extracting SIFT features and combining an RANSAC algorithm, and the image registration and splicing precision is remarkably improved; the splicing line is generated by adopting dynamic planning, and feathering processing is performed on the splicing edge, so that seam artifacts and brightness inconsistency caused by multi-strip splicing are effectively reduced, and the continuity and visual consistency of the spliced image are improved. The method is suitable for large-range remote sensing data fusion and application.

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing image processing, and in particular relates to a method for stitching vertical track scanning remote sensing satellite images. Background Art

[0002] Vertical track scanning imaging is a new type of satellite imaging method that combines wide width with high resolution. The vertical track scanning optical camera adopts a multi-strip stitching mode of a multi-CCD combination to obtain ultra-wide width images: that is, multiple linear array CCD detectors are stitched into a longer focal plane array and arranged along the direction of satellite flight. As the satellite moves in orbit, it uses a large-angle rotation scan of the optical mechanical device to achieve large-scale synchronous imaging on both sides of the track, thereby obtaining wide-field remote sensing images across multiple strips in a single pass. This multi-CCD multi-strip collaborative imaging mode breaks through the bottleneck of the limited width of a single CCD camera, enabling the satellite to significantly expand the coverage width while maintaining resolution. But the problem that follows is that the differences in the acquisition conditions of the images of each strip make stitching more difficult.

[0003] First, the CCD viewing angles and imaging geometries corresponding to different strips are different. Because the relative position and orientation of each CCD in the optical system vary slightly, and the satellite's attitude may change slightly during imaging, even for the same ground target, the image position and deformation in different strips are not exactly the same. This manifests itself as geometric inconsistency: the scales of the image's center area and edge strips may differ, distorting the shape of the ground objects. A wide-format vertical-track scanning image often exhibits regional scale changes and geometric distortion, with initial misalignment between substrips (subframes). Without sufficient geometric correction, these original subframes cannot be directly stitched together into a continuous, seamless, and internally accurate image of the entire scene. After stitching, misalignment may easily occur at the strip boundary.

[0004] Secondly, differences in exposure and radiation response between different strips can lead to image grayscale inconsistencies. Multiple CCD elements may have different sensitivity characteristics and dynamic ranges due to differences in manufacturing calibration. Furthermore, the acquisition times of each strip may vary slightly, and solar lighting conditions and atmospheric conditions may also vary. These factors lead to incomplete matching of brightness and hue between adjacent strips. Direct stitching often results in noticeable brightness differences or color casts at the seams. Without smooth transition processing, the visual consistency and quantitative accuracy of the stitched image will be affected.

[0005] In addition, under the conditions of large-sweep angle imaging, the content differences between the image strips are further exacerbated. In vertical-track sweep mode, when the camera's optical axis deviates from the vertical direction of the satellite track at a large angle (for example, scanning an area far from the sub-satellite point), the side-view effect of the ground objects is significant: some tall targets (such as buildings and mountains) may appear in side images in the side strips, while the top-view image is mainly presented in the strip directly below. Observing the ground objects from different perspectives will result in incomplete correspondence between the image content of adjacent strips. Even after position correction, the junction of the strips may still be difficult to seamlessly stitch together due to inconsistent content. The undulating terrain during large-field-of-view imaging also amplifies this difference - the projection positions of highlands and lowlands at different perspectives are more obviously different, requiring more complex correction.

[0006] In summary, the stitching of multi-CCD, multi-strip images in vertical-track scanning mode faces significant challenges due to geometric and radiometric inconsistencies: these include inter-strip positioning deviations, image distortion and scale differences, and discontinuities in brightness and color. These issues make high-quality image stitching extremely difficult. However, the quality of stitching has a significant impact on downstream remote sensing applications. The integrity and consistency of the stitched image are directly related to the reliability of subsequent information extraction, which not only interferes with the judgment of interpreters but also reduces the accuracy of algorithmic analysis. Therefore, achieving high-precision, seamless stitching of multi-strip remote sensing images from vertical-track scanning is crucial for ensuring the quality of results in large-scale remote sensing monitoring tasks (such as land cover mapping and disaster change detection), and is also a key issue that needs to be urgently addressed in current remote sensing image processing technology. Summary of the Invention

[0007] The purpose of the present invention is to provide a vertical-track scanning remote sensing satellite image stitching method to solve the many challenges faced by existing remote sensing image stitching methods in the fusion processing of vertical-track scanning multi-strip images, including positioning deviation between strips, image distortion and scale differences, and discontinuity of brightness and color.

[0008] To achieve the above object, the present invention provides the following technical solutions: The present invention provides a vertical track scanning remote sensing satellite image stitching method, the stitching method comprising the following steps: Step S1: Feature extraction and matching: extract and match features of multiple images with overlapping areas to obtain initial feature point pairs; Step S2: RANSAC matching optimization: For the initial feature point correspondence set, the matching results are robustly estimated to eliminate false matches, and the homography transformation matrix is ​​solved to align the relative positions of the images; Step S3: Optimal stitching line generation: Calculate the difference distribution between images in the overlapping area of ​​the images, and use a dynamic programming algorithm to search for an optimal stitching path passing through the overlapping area to generate the optimal stitching line; Step S4: Feathering processing of the seam area: along the optimal stitching line, pixel weighted fusion and feathering smoothing processing are performed on the edge area where the two images meet; Step S5: seamless image output: the multiple image data processed in the above steps are stitched together according to the optimal stitching line to output a seamlessly blended stitched image.

[0009] Furthermore, in another preferred embodiment, the above step S1 is specifically as follows: Step S11: extracting a large number of feature points from multiple images with overlapping areas using a scale-invariant feature transformation algorithm and calculating their feature description vectors; Step S12: Match the calculated feature description vectors between adjacent images to obtain initial feature point correspondence pairs.

[0010] Furthermore, in another preferred embodiment, the above step S2 is specifically as follows: Step S21: Use the Random Sampling Consensus Algorithm (RANSAC) to repeatedly randomly extract a small number of feature pairs for robust estimation, remove obvious mismatched outliers, and retain the set of inliers with good consistency; Step S22: Based on the filtered interior point correspondences, the homography transformation matrix or other spatial transformation parameters between the images are calculated to complete the precise geometric registration between the overlapping images and align the relative positions of the images.

[0011] Furthermore, in a preferred embodiment, the optimal stitching line selects a path with the smallest gradient change value to pass through the overlapping area.

[0012] Furthermore, in a preferred embodiment, the above-mentioned optimal stitching line serves as a boundary between different images, and the front and back images are stitched together at this path.

[0013] Furthermore, in another preferred embodiment, the above pixel weighted fusion is specifically as follows: A transition zone is established near the optimal stitching line, and the overlapping image pixels are weighted averaged or gradually blended to make the images gradually transition on both sides of the seam.

[0014] Furthermore, in another preferred embodiment, before generating the optimal stitching line, image resampling is performed, specifically: Taking the first strip image as the reference coordinate system, the other strip images are mapped to the reference coordinate system through the obtained homography matrix, and the bicubic interpolation algorithm is used for resampling during the mapping process.

[0015] The vertical track scanning remote sensing satellite image stitching method described in the present invention can be fully implemented using computer software. Therefore, correspondingly, the present invention also provides a vertical track scanning remote sensing satellite image stitching system, which includes a storage device, and the storage device is used to execute the vertical track scanning remote sensing satellite image stitching method described in the above preferred embodiment.

[0016] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the vertical track scanning remote sensing satellite image stitching method described in any one of the above preferred embodiments is executed.

[0017] The present invention also provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the vertical track scanning remote sensing satellite image stitching method described in any one of the above preferred embodiments.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. This paper proposes a method for stitching vertical-track scanning remote sensing satellite images. By extracting SIFT features and combining them with the RANSAC algorithm to obtain robust matching and homography transformation, the accuracy of image registration and stitching is significantly improved. Dynamic programming is also used to generate stitching lines and feathering is performed on the stitching edges, effectively reducing seam artifacts and brightness inconsistencies caused by stitching multiple strips, and improving the continuity and visual consistency of the stitched image.

[0019] Furthermore, before generating the optimal stitching line, a global resampling correction step based on an accurate geometric model and cubic spline interpolation effectively eliminates the geometric distortion caused by imaging, so that the output image has a consistent geometric reference in geographic space.

[0020] 2. The method proposed in the present invention takes into account both efficiency and accuracy. Through the image slicing parallel processing mechanism, it has high processing efficiency and good scalability, and is suitable for automatic batch processing of large-scale remote sensing images.

[0021] In summary, the splicing method proposed in the present invention has the advantages of high precision, high consistency and high efficiency compared with the existing technology.

[0022] The present invention processes multi-strip satellite images of rotary sweep imaging to achieve seamless splicing of remote sensing images, which is convenient for subsequent large-scale remote sensing data fusion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 This is a schematic diagram of the overall scheme of a vertical track scanning remote sensing satellite image stitching method proposed in the present invention; Figure 2 This is the large side-swing angle image stitching result of the present invention; Figure 3 is the mosaic result of the subsatellite point image according to the present invention; Figure 4 It is the optimal inlay line of the present invention; Figure 5 is a schematic diagram of the overlapping strips of the present invention; Figure 6 This is the SIFT feature point extraction process described in the present invention. DETAILED DESCRIPTION

[0025] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details that obscure the description of the present application.

[0026] The specific embodiments of the present invention are further described below in conjunction with the accompanying drawings. The following embodiments will help those skilled in the art further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that those skilled in the art may make various changes and improvements without departing from the scope of the present invention, and these are all within the scope of protection of the present invention.

[0027] Implementation Method 1: Combination Figure 1Explain this embodiment. Existing remote sensing image stitching methods face many challenges in the fusion processing of multiple strips of vertical track scanning images. This embodiment proposes improvement solutions for the following typical stitching problems: First, due to the differences in imaging geometry of multiple CCD strips and the slight disturbance of satellite attitude during the scanning process, adjacent images will have pixel-level or even sub-pixel-level translation, rotation and scale errors at the initial position, affecting the precise alignment of overlapping areas; second, under the condition of vertical track large side swing imaging, the viewing angles of different strips deviate from the main vertical direction, resulting in significant parallax, and the elevation The projection positions of undulating land features (such as buildings and mountains) in different strips are inconsistent, which increases the differences in image content and brings greater challenges to matching and fusion. Thirdly, there are often large areas of repeated or low-texture areas (such as forests and farmland) in remote sensing images. Traditional feature extraction and matching algorithms are prone to mismatching, resulting in instability of the estimated registration transformation, which affects the quality of subsequent stitching. Fourthly, because the stitching line often passes through the structure of the land features or areas with complex textures, it is difficult to completely eliminate the artifacts at the seams through simple fusion or smoothing, affecting the continuity of the stitching image and the professional application effect. The above problems cause the images obtained by traditional stitching to have defects such as position deviation and visual incoherence, which reduces the practicality of the stitched images. The purpose of this embodiment is to solve these problems, achieve high-precision registration and seamless stitching of multiple remote sensing images, eliminate seam artifacts and correct brightness differences, and produce high-quality fused images.

[0028] The present embodiment proposes a vertical track scanning remote sensing satellite image stitching method, such as Figure 1 As shown, the following steps are included: Step S1: Feature extraction and matching: extract and match features of multiple images with overlapping areas to obtain initial feature point pairs; Step S2: RANSAC matching optimization: For the initial feature point correspondence set, the matching results are robustly estimated to eliminate false matches, and the homography transformation matrix is ​​solved to align the relative positions of the images; Step S3: Optimal stitching line generation: Calculate the difference distribution between images in the overlapping area of ​​the images, and use a dynamic programming algorithm to search for an optimal stitching path passing through the overlapping area to generate the optimal stitching line; Step S4: Feathering processing of the seam area: along the optimal stitching line, pixel weighted fusion and feathering smoothing processing are performed on the edge area where the two images meet; Step S5: seamless image output: the multiple image data processed in the above steps are stitched together according to the optimal stitching line to output a seamlessly blended stitched image.

[0029] The vertical-track scanning remote sensing satellite image stitching method proposed in this embodiment significantly improves the accuracy and visual effect of vertical-track remote sensing image stitching by introducing robust feature matching and optimization algorithms, combined with optimal stitching line extraction and gradient fusion technology, and can generate high-quality wide-format remote sensing image products.

[0030] Implementation 2: This implementation is a specific description of the vertical track scanning remote sensing satellite image stitching method proposed in the above implementation 1. Step S1: Feature extraction and matching: extract and match features of multiple images with overlapping areas to obtain initial feature point pairs; Specifically: The Scale Invariant Feature Transform (SIFT) algorithm is used to extract a large number of feature points from overlapping images and calculate their feature description vectors. These feature descriptions are then matched between adjacent images to obtain initial feature point pairs. This extraction of rich, scale- and rotation-invariant feature information ensures robust matching between images.

[0031] Step S2: RANSAC matching optimization: For the initial feature point correspondence set, the matching results are robustly estimated to eliminate false matches, and the homography transformation matrix is ​​solved to align the relative positions of the images; Specifically: For the initial set of matching point pairs, the Random Sample Consensus (RANSAC) algorithm is used to robustly estimate the matching results. This algorithm repeatedly randomly extracts a small number of feature pairs for model estimation, eliminating obvious mismatched outliers and retaining only the set of inliers with good consistency. Based on the correspondences between these filtered inliers, the precise homography transformation matrix or other spatial transformation parameters between the images are calculated, thereby achieving precise geometric registration between overlapping images and aligning the relative positions of the images.

[0032] Step S3: Optimal stitching line generation: Calculate the difference distribution between images in the overlapping area of ​​the images, and use a dynamic programming algorithm to search for an optimal stitching path passing through the overlapping area to generate the optimal stitching line; Specifically: The difference distribution between the images is calculated within the overlapping region, and a dynamic programming algorithm is used to search for an optimal stitching path through the overlapping region. This stitching line selects a path with the smallest gradient change through the overlapping region, avoiding locations with significant structural differences or target objects as much as possible, thereby reducing visible differences in the stitching transition area. The resulting optimal stitching line serves as the boundary between the different images, and the previous and next images are stitched together at this path.

[0033] Step S4: Feathering processing of the seam area: along the optimal stitching line, pixel weighted fusion and feathering smoothing processing are performed on the edge area where the two images meet; Specifically: Along the selected stitching line, pixel-weighted fusion and feathering smoothing are performed on the edge where the two images meet. Specifically, a transition zone of a certain width is established near the stitching line, and the overlapping image pixels are weighted averaged or gradient blended to create a gradual transition between the two sides of the seam. Feathering effectively eliminates brightness and color inconsistencies between images, reducing the abrupt edge effects caused by stitching, and ensuring a natural, unobtrusive transition at the seam.

[0034] Step S5: seamless image output: the multiple image data processed in the above steps are stitched together according to the optimal stitching line to output a seamlessly blended stitched image.

[0035] Specifically: The multiple images, which have undergone the aforementioned registration transformation and seam optimization, are then stitched together along the defined stitching lines to produce a seamlessly blended image. The resulting image is visually coherent, with smooth transitions between overlapping areas and no noticeable brightness differences, meeting the quality requirements for large-scale seamless imagery in remote sensing applications.

[0036] In summary, the method proposed in this embodiment takes into account both efficiency and accuracy. Through the parallel processing mechanism of image slices, it has high processing efficiency and good scalability, and is suitable for automatic batch processing of large-scale remote sensing images. It has the following advantages: (1) By extracting SIFT features and combining them with the RANSAC algorithm to obtain robust matching and homography transformation, the accuracy of image registration and stitching is significantly improved; (2) Dynamic programming is used to generate stitching lines and feathering is performed on the stitching edges, which effectively reduces the seam artifacts and brightness inconsistency caused by multi-strip stitching, and improves the continuity and visual consistency of the stitched image; (3) Based on the precise geometric model and the global resampling correction step of cubic spline interpolation, the geometric distortion caused by imaging is effectively eliminated, so that the output image has a consistent geometric reference in geographic space.

[0037] Implementation Method 3: Combination Figures 2 to 6 This embodiment is to describe the actual operation of the vertical track scanning remote sensing satellite image stitching method described in the above embodiment; For the vertical track scanning remote sensing imaging system, its multi-CCD splicing form is shown in Figure 5. A single strip has 4096 rows of pixels along the track direction, and adjacent strips overlap by 200 pixels.

[0038] Step 1: Scan the remote sensing image for local overlapping vertical tracks and extract SIFT feature points in the image. The extraction process is as follows: Figure 6 shown.

[0039] In this embodiment, the SIFT algorithm is first used to extract the key points of the two images, and the gradient direction histogram is accumulated in the neighborhood of each key point to obtain a descriptor vector with a length of 128. and , and its Euclidean distance is defined as: (1) By calculating the distance between a descriptor and all candidate descriptors, the matching pair with the smallest distance is selected as the nearest neighbor matching to complete the preliminary feature point matching.

[0040] Step 2: In this embodiment, the RANSAC (Random Sampling Consensus) algorithm is used to eliminate false matches. Let the current model be the homography matrix , the original image point pass Mapping to get the projection point : (2) in, . Define The projection error of a point is: (3) like Less than the preset threshold , then the matching point pair is determined to be an inlier, otherwise it is an outlier. RANSAC randomly selects models by iteration and counts the number of inliers, and finally retains the model with the most inliers and its corresponding matching point set.

[0041] Step 3: Homography matrix estimation: using at least four sets of plane corresponding points Estimating the homography matrix In homogeneous coordinates, two sets of corresponding points satisfy: (4) Removing homogeneous factors and expanding, we can get two sets of linear equations: (5) Each set of corresponding points generates two linear constraints. Taking group points as an example, the augmented matrix form can be constructed: (6) in, The equations of the four point pairs can be formed. Homogeneous linear equations Solve the system of equations (usually normalized) by singular value decomposition (SVD) or least squares method ), you can get the normalized homography matrix .

[0042] Step 4: Image resampling and mosaic line generation: Using the first strip image as the reference coordinate system, map the other strip images to the reference coordinate system using the obtained homography matrix. For each image, a bicubic interpolation algorithm is used for resampling during the mapping process to ensure image data continuity and interpolation accuracy. In the overlapping area of ​​the images, a cost matrix (based on gradient information) for matching pixels is constructed, and a dynamic programming method is used to find the path with the minimum cumulative cost as the stitching seam (mosaic line), such as Figure 4 This stitching line ensures the smoothest transition between the two image bands in the overlapping area, avoiding misalignment and aliasing caused by direct stitching.

[0043] Step 5: Image cropping, stitching, and feathering: Based on the generated mosaic lines, each pair of strip images is cropped at the stitching line, and then the cropped images are stitched together along the mosaic lines to form a continuous large image. Since there may be differences in brightness and hue when shooting different strips, feathering is performed on the stitching edge area to eliminate the obvious boundaries on both sides of the stitching seam. The edge pixels are blended using a weighted average method to achieve a smooth transition in brightness and color between the two image areas. After feathering is completed, the final stitched image is output, as shown in the figure below. Figure 2 and Figure 3 As shown, Figure 2 and Figure 3 The stitching results of large side-swing angle images and subsatellite point images are respectively shown. It can be seen from the figures that the method of the present invention has good consistency in vision and radiometry, and there is no obvious stitching trace.

[0044] Embodiment 4: The vertical-track scanning remote sensing satellite image stitching method proposed in the above embodiment can be fully implemented using computer software. Therefore, correspondingly, this embodiment provides a vertical-track scanning remote sensing satellite image stitching system, the system comprising: A storage device for extracting and matching features of multiple images with overlapping areas to obtain initial feature point correspondences; A storage device for performing robust estimation of the matching results to eliminate false matches based on the initial feature point correspondence set, and performing homography transformation matrix calculation to align the relative positions of the images; A storage device for calculating the difference distribution between images in the overlapping area of ​​images, searching for an optimal stitching path through the overlapping area using a dynamic programming algorithm, and generating an optimal stitching line; A storage device for performing pixel weighted fusion and feathering smoothing processing on the intersection edge area of ​​the two images along the optimal stitching line; A storage device for splicing the multiple image data processed in the above steps according to the optimal splicing line to output a seamlessly merged spliced ​​image.

[0045] Embodiment 5: This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the vertical track scanning remote sensing satellite image stitching method described in any one of the above embodiments is executed.

[0046] Implementation 6. This implementation provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the vertical track scanning remote sensing satellite image stitching method described in any one of the above embodiments.

[0047] This embodiment provides a computer device. The hardware device in this part is a general model and is not shown in the form of a diagram. The system includes a processor and a memory, wherein the processor and the memory can be connected via a bus or other means. The memory is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs and modules, and corresponding program instructions / modules. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory, so as to realize the vertical track scanning remote sensing satellite image stitching method and steps in the above method embodiment.

[0048] The foregoing description is merely an embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of the claims.

Claims

1. A vertical track scanning remote sensing satellite image stitching method, characterized in that: The method is: S1: Extract and match features of multiple images with overlapping areas to obtain initial feature point pairs; S2: For the initial feature point correspondence set, perform robust estimation on the matching results to eliminate false matches, and solve the homography transformation matrix to align the relative positions of each image; S3: Calculate the difference distribution between images in the overlapping area of ​​the images, and use the dynamic programming algorithm to search for an optimal stitching path through the overlapping area to generate the optimal stitching line; S4: along the optimal stitching line, pixel weighted fusion and feathering smoothing are performed on the intersection edge area of ​​the two images; S5: The multiple image data processed in the above steps are stitched together according to the optimal stitching line to output a seamlessly fused stitched image.

2. The vertical track scanning remote sensing satellite image stitching method according to claim 1, characterized in that: S1 is specifically: S11: A scale-invariant feature transformation algorithm is used to extract a large number of feature points from multiple images with overlapping areas, and their feature description vectors are calculated; S12: Match the calculated feature description vectors between adjacent images to obtain initial feature point correspondence pairs.

3. The vertical track scanning remote sensing satellite image stitching method according to claim 1, characterized in that: S2 is specifically: S21: Use the random sampling consensus algorithm to repeatedly randomly extract a small number of feature pairs for robust estimation, remove obvious mismatched outliers, and retain the set of inliers with good consistency; S22: Based on the filtered internal point correspondence, the homography transformation matrix or other spatial transformation parameters between the images are calculated to complete the precise geometric registration between the overlapping images and align the relative positions of the images.

4. The vertical track scanning remote sensing satellite image stitching method according to claim 1, characterized in that: The optimal stitching path selects the path with the smallest gradient change value through the overlapping area.

5. The vertical track scanning remote sensing satellite image stitching method according to claim 4, characterized in that: The optimal stitching line serves as the boundary between different images, and the previous and next images are stitched together at this path.

6. The vertical track scanning remote sensing satellite image stitching method according to claim 1, characterized in that: Pixel weighted fusion is specifically as follows: A transition zone is established near the optimal stitching line, and the overlapping image pixels are weighted averaged or gradually blended to make the images gradually transition on both sides of the seam.

7. The vertical track scanning remote sensing satellite image stitching method according to claim 1, characterized in that: Before generating the optimal stitching line, image resampling is performed, specifically: Taking the first strip image as the reference coordinate system, the other strip images are mapped to the reference coordinate system through the obtained homography matrix, and the bicubic interpolation algorithm is used for resampling during the mapping process.

8. Vertical track scanning remote sensing satellite image stitching system, characterized by: The system comprises a storage device configured to execute the method of claim 1 .

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the vertical-track scanning remote sensing satellite image stitching method according to any one of claims 1 to 7.

10. A computer device, characterized in that: The device includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the vertical track scanning remote sensing satellite image stitching method according to any one of claims 1 to 7.

Images