Extraction method of large-area remote sensing image mosaic line based on superpixel segmentation

Abstract

The invention relates to the field of remote sensing image processing and analysis, and discloses a method for extracting a large-area remote sensing image mosaic line based on superpixel segmentation, which comprises the following steps of: 1, preprocessing an image; step 2, image overlapping processing; step 3, cutting and segmenting an overlapping region; 4, processing line vector data; 5, intersection point processing and path solving are carried out; step 6, shortest path solving and image segmentation are carried out; step 7, selecting an inlaying area; and step 8, subsequent processing: after the extraction of the mosaic line is completed, further smoothing the mosaic line, and correcting the color consistency of the image. By downsampling the image, the data volume involved in the processing process is reduced, so that the overall processing speed is improved, and particularly, when a large-area remote sensing image is processed, the method remarkably reduces the consumption of computing resources, so that the image processing is more efficient.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing and analysis, and in particular to a method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation. Background Art

[0002] In the field of remote sensing image processing, image stitching (also known as mosaicking) is an important step, especially in the processing and analysis of large-area images. By stitching multiple remote sensing images into a complete image, a broader perspective of geographic information can be provided, which can be applied to environmental monitoring, urban planning, agricultural management and other fields.

[0003] However, when processing large-area remote sensing images, traditional remote sensing image mosaicking methods require frame-by-frame alignment and splicing, which results in huge computational complexity and slow processing speed. This problem is particularly prominent for high-resolution images, which seriously limits their efficiency and practicality in practical applications. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for extracting mosaic lines of large-area remote sensing images based on superpixel segmentation, which solves the problem of slow processing speed of traditional remote sensing image mosaic methods when processing large-area remote sensing images.

[0005] To achieve the above objectives, the present invention is implemented by the following technical scheme: A method for extracting mosaic lines of large-area remote sensing images based on superpixel segmentation comprises the following steps:

[0006] Step 1: Image preprocessing: When processing large-area remote sensing images, the original images need to be preprocessed first. The resolution of the images is reduced by downsampling. Then, the effective area contours of the images are extracted so that the subsequent steps focus on the parts of the images that contain actual information and avoid processing irrelevant areas.

[0007] Step 2: Image overlap processing: By detecting the overlapping parts between images, determine the areas that need further processing, avoid processing irrelevant parts, and limit the mosaic processing to the areas that are actually needed;

[0008] Step 3: Overlapping area cropping and segmentation: After detecting the overlapping areas of the image, these overlapping parts need to be further cropped and superpixel segmentation is performed. The segmented image area is divided into multiple parts with similar features;

[0009] Step 4: Line vector data processing: vectorize the segmentation results, convert them into line vector data, simplify and optimize these vector lines, and remove unnecessary edge parts;

[0010] Step 5: Intersection processing and path solving: By calculating the intersection points of the image contours, the starting point and end point of the mosaic line are determined, and then the distance to each node in the line vector is calculated based on these points;

[0011] Step 6: Shortest path solution and image segmentation: Use the Dijkstra algorithm to solve the shortest path in the vectorized data, and segment the image according to the path so that the mosaic line is generated along the most suitable path in the image;

[0012] Step 7, mosaic area selection: After the image is segmented, it is necessary to select a suitable area for the final mosaic. By comparing and selecting the area with the largest area, it meets the expected mosaic conditions;

[0013] Step 8, follow-up processing: After the mosaic line extraction is completed, the mosaic line is further smoothed and the color consistency of the image is corrected.

[0014] Preferably, in step 1, the original remote sensing image is first downsampled using an appropriate downsampling algorithm. The selection of the downsampling ratio needs to be weighed according to the resolution of the image and the computing power of the processing equipment. The appropriate downsampling algorithm includes bilinear interpolation and bicubic interpolation. Then, an edge detection algorithm is used to extract the effective area contour of the image, marked as S1. The effective area refers to the area containing actual useful information. Black edges or invalid data are removed, and the extracted contour is morphologically processed to remove noise. The edge detection algorithm includes Canny edge detection, and the morphological processing includes dilation and erosion.

[0015] Preferably, in step 2, it is necessary to perform overlap detection on the current image and its subsequent images in turn to determine whether there is an overlap between their effective area contours S1, and then use geometric transformation to align the two images. This step needs to be assisted by an image feature point matching algorithm. If it is detected that there is an overlap between the two images, the contour of the overlapping area is extracted and marked as S2; if there is no overlap, the image is skipped. The geometric transformation includes affine transformation, and the image feature point matching algorithm includes SI FT and SURF.

[0016] Preferably, in step 3, the overlapping area cropping uses the overlapping contour S2 to crop the original image, and the image data of the overlapping part is extracted and marked as R1; the superpixel segmentation uses the SLIC superpixel segmentation algorithm to segment the cropped image R1 to obtain the segmentation result R2; the SLIC algorithm clusters the image pixels into superpixels in an iterative manner, and each superpixel represents an area in the image with similar color or texture features; when implementing the segmentation, the parameters of the SLIC algorithm are adjusted to control the detailedness of the segmentation.

[0017] Preferably, in step 4, the result R2 after superpixel segmentation is converted into line vector data, marked as S3, and then the Douglas-Peucker algorithm is used to simplify the line vector of S3; the line vector data S3 is thinned, the edge part of the road network data is deleted, and the optimized line vector data is marked as S4.

[0018] Preferably, in step 5, the intersection points of the effective area contours of the two images are calculated. If the number of intersection points is equal to 2, they are used as the starting point and the end point respectively. If the number of intersection points is greater than 2, the distance between each two points is calculated, and the two points with the largest distance are used as the starting point and the end point. For each node in S4, the distance between them and the starting point and the end point is calculated respectively, and the distance information is saved in lists L1 and L2.

[0019] Preferably, in step 6, the Dijkstra algorithm is used to solve the shortest paths between nodes at the same position in L1 and L2 in the road network data of S4 in sequence; if a feasible path is not found at a certain position, try to use the nodes at the next position in L1 and L2 to re-solve until a feasible path is found; and the original image is segmented according to the obtained shortest path, and the image is segmented into two parts, D1 and D2.

[0020] Preferably, in step 7, for the segmented regions D1 and D2, the region with the largest area is first selected. If the selected region is included in S2, another region is selected as the final mosaic region.

[0021] Preferably, in step 8, the finally extracted mosaic lines are smoothed using a B-spline curve smoothing algorithm, and color consistency correction is performed using a histogram matching or linear transformation method.

[0022] The present invention provides a method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation. It has the following beneficial effects:

[0023] 1. The present invention reduces the amount of data involved in the processing process by downsampling the image, thereby improving the overall processing speed. Especially when processing large-area remote sensing images, this method significantly reduces the consumption of computing resources and makes image processing more efficient.

[0024] 2. The present invention uses the SLIC superpixel segmentation algorithm to segment images more accurately, dividing complex images into regions with similar features. This fine segmentation improves the accuracy of mosaic line extraction and ensures that the mosaic lines can accurately reflect the actual features in the image.

[0025] 3. The present invention avoids unnecessary image processing and reduces processing errors caused by misjudgment of overlapping areas through accurate detection and cutting of overlapping areas. Accurate contour extraction and path calculation also reduce errors in processing, making the final result more reliable.

[0026] 4. The present invention makes the mosaic line more concise and smooth through line vectorization and thinning processing, avoiding the appearance of redundant lines. Coupled with the subsequent smoothing processing, the generated mosaic line is more visually beautiful and meets the requirements of high-quality image stitching. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of the method for extracting mosaic lines of large-area remote sensing images based on superpixel segmentation of the present invention;

[0028] Figure 2 It is a schematic diagram of superpixel segmentation of the present invention;

[0029] Figure 3 It is a schematic diagram of the overlapping area cutting and inlaying lines of the present invention. DETAILED DESCRIPTION

[0030] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] Please refer to the attached Figure 1-Figure 3 The embodiment of the present invention provides a method for extracting mosaic lines of a large-area remote sensing image based on superpixel segmentation, comprising the following steps:

[0032] Step 1: Image preprocessing: When processing large-area remote sensing images, the original images must first be preprocessed to improve processing efficiency. By downsampling, the resolution of the image is reduced, thereby increasing the calculation speed. Next, the effective area contour of the image is extracted to ensure that subsequent steps focus on the part of the image that contains actual information and avoid processing irrelevant areas. This step lays the foundation for subsequent segmentation and mosaic line extraction;

[0033] Step 2: Image overlap processing: In order to successfully mosaic multiple images together, their overlapping areas need to be identified. By detecting the overlapping parts between the images, the areas that need further processing are determined, and irrelevant parts are avoided. This step ensures that the mosaic processing is limited to the areas that are actually needed, improves processing efficiency, and reduces the amount of calculation;

[0034] Step 3: Overlapping area cropping and segmentation: After detecting the overlapping areas of the image, it is necessary to further crop these overlapping parts and perform superpixel segmentation. The segmented image area is divided into multiple parts with similar features, which is convenient for subsequent vectorization processing. Through this segmentation method, the feature areas in the image can be defined more accurately, thereby improving the accuracy of mosaic line extraction;

[0035] Step 4: Line vector data processing: vectorize the segmentation results, convert them into line vector data, simplify and optimize these vector lines, and remove unnecessary edge parts. The importance of this step is that it provides a concise and accurate data basis for subsequent path solving and image segmentation, ensuring the final quality of the mosaic line;

[0036] Step 5, intersection processing and path solving: By calculating the intersection points of the image contours, determine the starting and ending points of the mosaic line, and then calculate the distance to each node in the line vector based on these points. The purpose of this step is to provide reference points for subsequent path solving. By reasonably selecting the starting and ending points, the path calculation process can be optimized and the generation of erroneous paths can be reduced;

[0037] Step 6, shortest path solution and image segmentation: Use the Dijkstra algorithm to solve the shortest path in the vectorized data, and segment the image based on the path. Ensure that the mosaic line is generated along the most suitable path in the image to reduce mosaic errors. This step is the core of mosaic line extraction. By optimizing the path calculation, the accuracy of image segmentation can be improved;

[0038] Step 7, mosaic area selection: After the image is segmented, it is necessary to select a suitable area for the final mosaic. By comparing and selecting the largest area and ensuring that it meets the expected mosaic conditions, it can ensure that the mosaicked image has a good effect. This step determines the final mosaic area, which affects the overall visual effect and practical application effect;

[0039] Step 8, follow-up processing: After completing the mosaic line extraction, you can further smooth the mosaic line and correct the color consistency of the image to ensure the beauty and consistency of the final result. This step is to optimize the mosaic result, ensuring that the image looks natural and seamless after stitching, and the color transition is smooth, which is suitable for scenes with high requirements for image quality in practical applications.

[0040] In step 1, the original remote sensing image is downsampled using an appropriate downsampling algorithm (bilinear interpolation, bicubic interpolation). This step is intended to reduce the resolution of the image, thereby increasing the subsequent processing speed.

[0041] The selection of the downsampling ratio needs to be weighed against the image resolution and the computing power of the processing device to ensure a balance between processing speed and accuracy.

[0042] The edge detection algorithm (Canny edge detection) is used to extract the effective area contour of the image, which is marked as S1. The effective area usually refers to the area containing actual useful information, removing the black edges or invalid data.

[0043] The extracted contours are subjected to morphological processing (dilation and erosion) to remove noise and ensure the integrity and coherence of the contours.

[0044] In step 2, overlap detection is performed on the current image and its subsequent images in sequence to determine whether there is any overlap in their effective area contours S1.

[0045] Use geometric transformation (affine transformation) to align the two images to ensure the accuracy of the overlapping parts. This step may require the assistance of image feature point matching algorithms (SIFT, SURF).

[0046] If it is detected that there is an overlap between the two images, the outline of the overlapping area is extracted and marked as S2; if there is no overlap, the image is skipped.

[0047] In step 3, the overlapping contour S2 is used to crop the original image, and the image data of the overlapping part is extracted and marked as R1. The cropping process should try to maintain the geometric shape of the image to ensure the accuracy of subsequent processing.

[0048] The cropped image R1 should be saved as a high-precision image to ensure that details are preserved in the subsequent superpixel segmentation process.

[0049] The cropped image R1 is segmented using the SLIC (Simple Linear Interference Definition) superpixel segmentation algorithm to obtain the segmentation result R2.

[0050] The SLIC algorithm iteratively clusters image pixels into superpixels, each of which represents an area in the image with similar color or texture features.

[0051] When implementing segmentation, the parameters of the SLIC algorithm (number of superpixels, compactness factor) can be adjusted to control the level of detail of the segmentation and ensure that the boundaries of the segmented area are consistent with the actual image features.

[0052] In step 4, the result R2 after superpixel segmentation is converted into line vector data, which is marked as S3. Line vectorization is to convert the edge of the segmented area into vector lines for subsequent path calculation and thinning processing.

[0053] The Douglas-Peucker algorithm is used to simplify the line vector of S3, reducing the number of line segments while maintaining the main features of the contour.

[0054] The line vector data S3 is thinned, the edge part of the road network data is deleted, and the optimized line vector data is marked as S4.

[0055] During the thinning process, it is necessary to ensure that key geographical features (roads, rivers) are retained and those small edges that are not relevant to the mosaic are removed.

[0056] In step 5, the intersection points of the effective area contours of the two images are calculated. If the number of intersection points is equal to 2, they are used as the starting point and the end point respectively.

[0057] If the number of intersection points is more than 2, calculate the distance between every two points and take the two points with the largest distance as the starting point and the end point. This process can use Euclidean distance or other appropriate distance measurement methods.

[0058] For each node in S4, calculate the distance between them and the starting point and the end point respectively, and save the distance information into lists L1 and L2.

[0059] Arrange the distances in L1 and L2 in ascending order for subsequent path solving.

[0060] In step 6, the Dijkstra algorithm is used to solve the shortest paths between nodes at the same position in L1 and L2 in the road network data of S4.

[0061] The Dijkstra algorithm gradually expands the path through a greedy strategy to ensure that the found path has the shortest distance.

[0062] If a feasible path is not found at a certain position, try to solve again using the nodes at the next position in L1 and L2 until a feasible path is found.

[0063] The original image is segmented according to the obtained shortest path, and the image is divided into two parts, D1 and D2.

[0064] It is necessary to ensure that the segmentation path can effectively distinguish the areas that need to be mosaicked and the non-mosaiced areas in the image to avoid mis-segmentation.

[0065] In step 7, for the segmented regions D1 and D2, the region with the largest area is first selected.

[0066] If the selected area is included in S2, another area is selected as the final mosaic area.

[0067] When selecting a mosaic area, the geometric shape and actual geographical features of the area must also be considered to ensure that the mosaic effect is natural and beautiful.

[0068] In step 8, the final extracted mosaic lines are smoothed to avoid the lines being too harsh or jagged. B-spline curves or other smoothing algorithms may be used for processing.

[0069] After the images are mosaicked, color consistency correction is performed to ensure that the mosaicked images have no obvious color difference visually. Color correction can be performed using histogram matching or linear transformation methods.

[0070] The following are explanations of some of the professional names and technical means mentioned in the method:

[0071] 1. Downsampling

[0072] Explanation: Downsampling refers to reducing the resolution or sampling rate of an image to reduce the amount of image data in order to increase processing speed. In remote sensing image processing, it is often used to reduce the size of an image before preliminary analysis or processing, thereby reducing the computational burden.

[0073] Application: By reducing the resolution of an image, preliminary processing can be done more quickly without having to process the entire high-resolution image at once.

[0074] 2.Effective Area Contour

[0075] Explanation: The effective area contour refers to the boundary of the area in the image that contains actual useful information. It is usually used to remove invalid edge areas, such as black edges, blank areas, etc. In remote sensing images, extracting the effective area contour can focus the processing on the actual area of ​​interest.

[0076] Application: In remote sensing image processing, the effective area contour helps to remove unnecessary background areas, thereby improving processing efficiency and accuracy.

[0077] 3. Edge Detection Algorithm

[0078] Explanation: Edge detection is an image processing technique used to identify areas with large brightness changes in an image, i.e., edges. Commonly used edge detection algorithms include Canny, Sobel, Prewitt, etc. These algorithms detect edges by calculating the gradient of pixel values ​​in an image.

[0079] Application: When extracting the effective area contour, edge detection can help identify the actual boundary in the image for subsequent contour extraction and processing.

[0080] 4. Geometric Transformation

[0081] Explanation: Geometric transformation is the adjustment of the position, angle, size, etc. of the image so that the image can be aligned in space. Common geometric transformations include translation, rotation, scaling, affine transformation, etc. In the image registration process, geometric transformation is used to adjust different images to the same spatial reference system.

[0082] Application: In the overlap detection of remote sensing images, the overlapping areas of different images are aligned through geometric transformation to ensure the accuracy of the overlapping parts.

[0083] 5. SLIC superpixel segmentation algorithm (Simple Line near Iterative C ustering)

[0084] Explanation: SLIC is a superpixel segmentation algorithm that divides an image into multiple superpixel regions by clustering. Each superpixel region contains pixels with similar color or texture features, thereby simplifying the structure of the image.

[0085] Application: In remote sensing image processing, the SLIC superpixel segmentation algorithm can segment complex images into multiple easy-to-process regions, which helps to extract more accurate image features.

[0086] 6. Line Vectorization

[0087] Explanation: Line vectorization is the process of converting the boundaries or contours in an image into vector form (i.e., a graphic composed of points and lines). Vector data has higher accuracy than raster data and is easier to perform geometric analysis and path calculation.

[0088] Application: In image processing, line vectorization is used to convert the area boundaries after superpixel segmentation into lines for subsequent path solving and mosaic line extraction.

[0089] 7. Douglas-Peucker algorithm

[0090] Explanation: Douglas-Peucker algorithm is an algorithm for simplifying curves or polygons by recursively deleting points on the curve or polygon and retaining key points to approximate the original shape. This method can reduce the number of data points but still retain the main features of the curve or contour.

[0091] Application: In online vector data processing, the Douglas-Peucker algorithm is used to simplify vector lines, thereby reducing processing complexity while maintaining the integrity of key geographic features.

[0092] 8. Dijkstra algorithm

[0093] Explanation: Dijkstra's algorithm is an algorithm for calculating the shortest path between two points in a graph. It finds the shortest path from the starting point to the end point by gradually expanding the known shortest path. This algorithm is widely used in network routing, path planning in geographic information systems and other fields.

[0094] Application: In image mosaic processing, the Dijkstra algorithm is used to find the shortest path between the starting point and the end point in the online vector data, so as to determine the best route for image segmentation.

[0095] 9. B-Spline Curve

[0096] Explanation: B-spline curve is a smooth curve used to represent complex curve shapes. It is generated by a set of control points and is more flexible and smoother than other curve representation methods (such as Bezier curves).

[0097] Application: In the subsequent smoothing process of the mosaic line, the B-spline curve can be used to smooth the line, reduce the jagged edge phenomenon, and make the mosaic line look more natural.

[0098] 10. Histogram Matching

[0099] Explanation: Histogram matching is an image processing technique used to adjust the color distribution of an image to match the histogram of a reference image. Color consistency is achieved by adjusting the grayscale or color histogram of the image.

[0100] Application: In the mosaicked image processing, histogram matching can be used to adjust the color difference between different images to ensure that there is no obvious difference in color between the mosaicked images.

[0101] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation, characterized in that: The following steps are involved: Step 1: Image preprocessing: When processing large-area remote sensing images, the original images need to be preprocessed first. The resolution of the images is reduced by downsampling. Then, the effective area contours of the images are extracted so that the subsequent steps focus on the parts of the images that contain actual information and avoid processing irrelevant areas. Step 2: Image overlap processing: By detecting the overlapping parts between images, determine the areas that need further processing, avoid processing irrelevant parts, and limit the mosaic processing to the areas that are actually needed; Step 3: Overlapping area cropping and segmentation: After detecting the overlapping areas of the image, these overlapping parts need to be further cropped and superpixel segmentation is performed. The segmented image area is divided into multiple parts with similar features; Step 4: Line vector data processing: vectorize the segmentation results, convert them into line vector data, simplify and optimize these vector lines, and remove unnecessary edge parts; Step 5: Intersection processing and path solving: By calculating the intersection points of the image contours, the starting point and end point of the mosaic line are determined, and then the distance to each node in the line vector is calculated based on these points; Step 6: Shortest path solution and image segmentation: Use the Dijkstra algorithm to solve the shortest path in the vectorized data, and segment the image based on the path so that the mosaic line is generated along the most suitable path in the image; Step 7, mosaic area selection: After the image is segmented, it is necessary to select a suitable area for the final mosaic. By comparing and selecting the area with the largest area, it meets the expected mosaic conditions; Step 8, follow-up processing: After the mosaic line extraction is completed, the mosaic line is further smoothed and the color consistency of the image is corrected.

2. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In the step 1, firstly, an appropriate downsampling algorithm is used to downsample the original remote sensing image. The selection of the downsampling ratio needs to be weighed according to the resolution of the image and the computing power of the processing equipment. The appropriate downsampling algorithm includes bilinear interpolation and bicubic interpolation. Then, an edge detection algorithm is used to extract the effective area contour of the image, marked as S1. The effective area refers to the area containing actual useful information. Black edges or invalid data are removed, and the extracted contour is morphologically processed to remove noise. The edge detection algorithm includes Canny edge detection, and the morphological processing includes dilation and erosion.

3. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 2, it is necessary to perform overlap detection on the current image and its subsequent images in turn to determine whether there is any overlap in their effective area contours S1, and then use geometric transformation to align the two images. This step needs to be assisted by an image feature point matching algorithm. If it is detected that there is an overlap between the two images, the contour of the overlapping area is extracted and marked as S2; if there is no overlap, the image is skipped. The geometric transformation includes affine transformation, and the image feature point matching algorithm includes SIFT and SURF.

4. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 3, the overlapping area cropping uses the overlapping contour S2 to crop the original image, and the image data of the overlapping part is extracted and marked as R1; the superpixel segmentation uses the SLIC superpixel segmentation algorithm to segment the cropped image R1 to obtain the segmentation result R2; the SLIC algorithm clusters the image pixels into superpixels in an iterative manner, and each superpixel represents an area in the image with similar color or texture features; when implementing the segmentation, the parameters of the SLIC algorithm are adjusted to control the detailed degree of the segmentation.

5. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In the step 4, the result R2 after superpixel segmentation is converted into line vector data, marked as S3, and then the Douglas-Peucker algorithm is used to simplify the line vector of S3; the line vector data S3 is thinned, the edge part of the road network data is deleted, and the optimized line vector data is marked as S4.

6. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 5, the intersection points of the effective area contours of the two images are calculated. If the number of intersection points is equal to 2, they are used as the starting point and the end point respectively. If the number of intersection points is more than 2, the distance between every two points is calculated, and the two points with the largest distance are used as the starting point and the end point. For each node in S4, calculate the distance between them and the starting point and the end point respectively, and save the distance information into lists L1 and L2.

7. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 6, the Dijkstra algorithm is used to solve the shortest paths between nodes at the same position in L1 and L2 in the road network data of S4 in turn; if a feasible path is not found at a certain position, try to use the nodes at the next position in L1 and L2 to solve again until a feasible path is found; the original image is segmented according to the obtained shortest path, and the image is segmented into two parts, D1 and D2.

8. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 7, for the segmented regions D1 and D2, the region with the largest area is first selected. If the selected region is included in S2, another region is selected as the final mosaic region.

9. The method for extracting mosaic lines from large-area remote sensing images based on superpixel segmentation according to claim 1, characterized in that: In step 8, the finally extracted mosaic lines are smoothed using a B-spline curve smoothing algorithm, and color consistency correction is performed using a histogram matching or linear transformation method.

Images