A post-processing method for road extraction of remote sensing images

Abstract

To address the limitations of existing technologies, this invention proposes a post-processing method for road extraction from remote sensing images. The solution of this invention incorporates the post-processing of road extraction results into the feature extraction process, which to a certain extent improves the connectivity and integrity of road information, preserves and restores the linear information of roads to the maximum extent, and realizes the complete extraction of road information from remote sensing images.

Description

Technical Field

[0001] This invention relates to the field of automated remote sensing image processing technology, and more specifically, to a post-processing method for road extraction from remote sensing images. Background Technology

[0002] Remote sensing technology, as a multi-scale, wide-coverage Earth observation technique, provides powerful data and technical services for social development by rapidly collecting geographic information within the research area and combining it with relevant analysis. With the continuous deepening of remote sensing research, remote sensing imagery, as an indispensable data support for scientific research and urban development planning, is widely used in various fields. Remote sensing imagery is characterized by short round-trip times and high spatial resolution, containing a large amount of spatial information about ground features, providing a rapid means of collecting and acquiring geospatial information, and offering excellent data support for urban management and regional planning.

[0003] As an important geographical element and urban infrastructure, roads play a vital role in spatial planning, urban construction, traffic navigation, and decision-making. In particular, the distribution, undulation, and connectivity information of roads are crucial in urban construction and traffic navigation. By extracting road information from remote sensing images through deep learning combined with image processing algorithms, we can provide practical, rapid, and effective data and technical support for planning and strategic deployment.

[0004] A Chinese invention patent published on October 23, 2020, describes a post-processing method for road network extraction from remote sensing images used in scene restoration. This method uses vectorization to further restore the machine-extracted binary road network image to a real scene, thereby calculating the road network width and repairing unidentified breaks in the road network image. However, the aforementioned prior art still has limitations in processing and repairing noisy isolated patches and layer holes generated during the road feature information extraction process. Summary of the Invention

[0005] To address the limitations of existing technologies, this invention proposes a post-processing method for road extraction from remote sensing images. The technical solution adopted by this invention is as follows:

[0006] A post-processing method for road extraction from remote sensing images includes the following steps:

[0007] S1, acquire remote sensing images, divide the remote sensing images into sections according to a preset cropping overlap; perform road segmentation on the sectioning results to obtain a predicted image containing roads and background;

[0008] S2, refine the predicted image, retaining the skeleton lines containing complete road structure information;

[0009] S3, perform eight-neighborhood analysis on the points on the skeleton line to extract the breakpoints on the skeleton line.

[0010] S4. By performing connected component analysis on the breakpoint, irrelevant data in the breakpoint is filtered out.

[0011] S5, perform eight-neighborhood analysis on the filtering results of step S4 to exclude breakpoints that are interfered with by redundant pixels and form a set of points to be fitted.

[0012] S6, Fit the set of fitted points, and draw the corresponding curve in the predicted image based on the fitting result to form complete road information.

[0013] Compared with existing technologies, the post-processing method for road extraction from remote sensing images of the present invention incorporates the post-processing of road extraction results into the feature extraction process, which to a certain extent improves the connectivity and integrity of road information, preserves and restores the linear information of roads to the maximum extent, and realizes the complete extraction of road information from remote sensing images.

[0014] As a preferred embodiment, in step S2, the predicted image is first subjected to the following repair process, and then the predicted image is refined:

[0015] After the predicted image undergoes grayscale, binarization, and dilation / erosion preprocessing, the independent patches in the predicted image are searched, and the area of ​​each independent patch is counted. Independent patches with an area smaller than a preset area threshold are removed from the predicted image. Holes in the predicted image layer are filled.

[0016] As a preferred embodiment, step S3 includes the following process:

[0017] Calculate the sum of pixels in eight directions around a point on the skeleton line, and take the point where the sum of pixels is 1 as the break point on the skeleton line.

[0018] Furthermore, step S4 includes the following process:

[0019] Connectivity analysis is performed on the breakpoints to divide breakpoints belonging to the same independent plot into coplanar point groups; the minimum Euclidean distance between each coplanar point group and the breakpoints outside the coplanar point group is obtained by calculating the Euclidean distance between each coplanar point group. d Iterate through the minimum Euclidean distance Min. d :

[0020] If the minimum value of the Euclidean distance is Min d If the distance is greater than the preset distance threshold P, then the Euclidean distance is considered to be minimum (Min). dThe corresponding coplanar point set does not contain any points suitable for fitting, and is therefore irrelevant data; if the minimum Euclidean distance is Min... d If the distance is less than the distance threshold P, then the minimum Euclidean distance, Min, is used. d The two corresponding breakpoints are stored as one element in the filter point group.

[0021] Furthermore, in step S5, for each element in the filtering point group, the following process is included:

[0022] The two breakpoints stored in the element are used as center points for eight-neighbor analysis: if the sum of the pixel values ​​of the eight neighbors is equal to 1, the coordinates [k,l] of the points with adjacent pixel values ​​of 1 are stored in the set of points to be fitted.

[0023] Eight-neighbor analysis is performed with the point coordinates [k,l] as the center point: if the sum of the eight neighboring pixels of the point coordinates [k,l] is 2, then the adjacent unstored point coordinates [k+1,l] of the point coordinates [k,l] are stored in the set of points to be fitted; if the sum of the eight neighboring pixels of the point coordinates [k,l] is greater than 2, then it is determined that there are redundant pixels interfering with the point coordinates [k,l].

[0024] As a preferred embodiment, in step S6, multivariate function fitting is performed according to the following formula:

[0025] F(x)=Ax 3 +Bx 2 +Cx+D.

[0026] Furthermore, the cropping overlap is 100 pixels; the area threshold is 70*70.

[0027] This invention also includes the following:

[0028] A post-processing system for road extraction from remote sensing images includes, in sequence, a predicted image acquisition module, a skeleton line acquisition module, a breakpoint extraction module, an irrelevant data filtering module, a point set acquisition module, and a fitting and plotting module; wherein:

[0029] The predicted image acquisition module is used to acquire remote sensing images, divide the remote sensing images into frames according to a preset cropping overlap degree, and perform road segmentation on the frame division results to obtain a predicted image containing roads and background.

[0030] The skeleton line acquisition module is used to refine the predicted image and retain skeleton lines containing complete road structure information.

[0031] The breakpoint extraction module is used to perform eight-neighbor analysis on the points on the skeleton line to extract the breakpoints on the skeleton line.

[0032] The irrelevant data filtering module is used to filter out irrelevant data in the breakpoint by performing connected component analysis on the breakpoint.

[0033] The module for obtaining the set of points to be fitted is used to perform eight-neighbor analysis on the filtering results of the irrelevant data filtering module, and to exclude the breakpoints that are interfered with by redundant pixels to form the set of points to be fitted.

[0034] The fitting and drawing module is used to fit the set of fitting points and draw the corresponding curves in the predicted image based on the fitting results to form complete road information.

[0035] Compared with existing technologies, the post-processing system for road extraction from remote sensing images of the present invention incorporates the post-processing of road extraction results into the feature extraction process, which to a certain extent improves the connectivity and integrity of road information, preserves and restores the linear information of roads to the maximum extent, and realizes the complete extraction of road information from remote sensing images.

[0036] A storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the post-processing method for road extraction from remote sensing images as described above.

[0037] A computer device includes a storage medium, a processor, and a computer program stored in the storage medium and executable by the processor, wherein the computer program, when executed by the processor, implements the steps of the post-processing method for road extraction from remote sensing images as described above. Attached Figure Description

[0038] Figure 1 This is a schematic flowchart of a post-processing method for road extraction from remote sensing images provided in Embodiment 1 of the present invention;

[0039] Figure 2 This is a schematic diagram illustrating the principle of the post-processing method for road extraction from remote sensing images provided in Embodiment 1 of the present invention.

[0040] Figure 3 This is a schematic diagram of overlapping cutting in an embodiment of the present invention;

[0041] Figure 4 This is an example image of the predicted image according to an embodiment of the present invention;

[0042] Figure 5 This is an example diagram showing the road segment skeleton line extraction results according to an embodiment of the present invention;

[0043] Figure 6 This is a schematic diagram of the eight breakpoint regions in an embodiment of the present invention;

[0044] Figure 7 This is an example diagram of the breakpoint extraction results in an embodiment of the present invention;

[0045] Figure 8 This is an example of a set of points to be fitted after being screened in an embodiment of the present invention;

[0046] Figure 9 This is a schematic diagram of multivariate function fitting according to an embodiment of the present invention;

[0047] Figure 10 Line segment fitting and original image diagram for embodiments of the present invention;

[0048] Figure 11 This is a schematic diagram of a post-processing system for road extraction from remote sensing images provided in Embodiment 2 of the present invention. Detailed Implementation

[0049] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0050] It should be understood that the described embodiments are merely some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.

[0051] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0052] In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims. In the description of this application, it should be understood that the terms "first," "second," "third," etc., are used only to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0053] Furthermore, in the description of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. The invention will be further described below with reference to the accompanying drawings and embodiments.

[0054] To address the limitations of existing technologies, this embodiment provides a technical solution. The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0055] Example 1

[0056] For a post-processing method for road extraction from remote sensing images, please refer to [link / reference]. Figure 1 as well as Figure 2 This includes the following steps:

[0057] S1, acquire remote sensing images, divide the remote sensing images into sections according to a preset cropping overlap; perform road segmentation on the sectioning results to obtain a predicted image containing roads and background;

[0058] S2, refine the predicted image, retaining the skeleton lines containing complete road structure information;

[0059] S3, perform eight-neighborhood analysis on the points on the skeleton line to extract the breakpoints on the skeleton line.

[0060] S4. By performing connected component analysis on the breakpoint, irrelevant data in the breakpoint is filtered out.

[0061] S5, perform eight-neighborhood analysis on the filtering results of step S4 to exclude breakpoints that are interfered with by redundant pixels and form a set of points to be fitted.

[0062] S6, Fit the set of fitted points, and draw the corresponding curve in the predicted image based on the fitting result to form complete road information.

[0063] Compared with existing technologies, the post-processing method for road extraction from remote sensing images of the present invention incorporates the post-processing of road extraction results into the feature extraction process, which to a certain extent improves the connectivity and integrity of road information, preserves and restores the linear information of roads to the maximum extent, and realizes the complete extraction of road information from remote sensing images.

[0064] The remote sensing image used in this embodiment is a Google Earth high-resolution remote sensing image at a scale of 1:10000, with a size of 11761*7767 and a resolution of 0.5m. The image contains three bands: red (R), green (G), and blue (B). The main problems that may arise when using remote sensing image data for road extraction are: (1) Ground shadows formed by buildings, clouds, and trees near the road may interfere with road extraction. (2) Trees at both ends of the road may cover the original road layer, causing breaks in the extraction process.

[0065] Specifically, in step S1, the overlap can be set to 100 to perform image framing with 100 pixels of overlap (e.g., ...). Figure 3 As shown in the image, pixel overlap performs secondary extraction of image edge information, which enhances the extraction of road edge information. The image segmentation results are then input into a deep learning network for road segmentation, resulting in an image containing both roads and background (e.g., ...). Figure 4 (As shown).

[0066] In a preferred embodiment, in step S2, the predicted image is first subjected to the following restoration process, and then the predicted image is refined:

[0067] After the predicted image undergoes grayscale, binarization, and dilation / erosion preprocessing, the independent patches in the predicted image are searched, and the area of ​​each independent patch is counted. Independent patches with an area smaller than a preset area threshold are removed from the predicted image. Holes in the predicted image layer are filled.

[0068] Specifically, the predicted image is searched and statistically analyzed. Patches without overlapping parts are assigned corresponding numbers (counts[m], where m = 0, ..., k). Simultaneously, the patch area is calculated. An area threshold can be set as image width / 70 * height / 70; patches with areas smaller than this threshold are removed, retaining other feature patches. Then, the noise-removed predicted image undergoes secondary processing to fill in holes in the image layers, ensuring data integrity.

[0069] In step S2, since the road has a certain width and the area near the breakpoint is composed of a series of adjacent pixels, this embodiment refines the road to ensure data operability. While preserving the shape of the line segments, it retains the most basic skeleton features (which can be compared). Figure 4 and Figure 5 ).

[0070] In a preferred embodiment, step S3 includes the following process:

[0071] Calculate the sum of pixels in eight directions around a point on the skeleton line, and take the point where the sum of pixels is 1 as the break point on the skeleton line.

[0072] Specifically, this step involves classifying the points on the skeleton line. For a pixel k[i,j], its eight surrounding directions are: top left [i-1,j-1], bottom left [i-1,j], top right [i+1,j-1], bottom right [i+1,j+1], top [i,j-1], bottom [i,j+1], left [i-1,j], and right [i+1,j].

[0073] Sum=[i+1,j-1]+[i+1,j+1]+[i,j-1]+[i,j+1]+[i-1,j]+[i+1,j]+[i-1,j-1]+[i-1,j] (1)

[0074] If the pixel values ​​sum = 0 in all eight directions, the center point is an isolated point; if sum = 1, the center point is a breakpoint (e.g., ...). Figure 6 and Figure 7 (as shown); if sum>1, then the center point is an internal point. Perform a sliding search on the image, retain all breakpoints, and form a breakpoint group point[n] (n=0,...,k).

[0075] Furthermore, step S4 includes the following process:

[0076] Connectivity analysis is performed on the breakpoints to divide breakpoints belonging to the same independent plot into coplanar point groups; the minimum Euclidean distance between each coplanar point group and the breakpoints outside the coplanar point group is obtained by calculating the Euclidean distance between each coplanar point group. d Iterate through the minimum Euclidean distance Min. d :

[0077] If the minimum value of the Euclidean distance is Min d If the distance is greater than the preset distance threshold P, then the Euclidean distance is considered to be minimum (Min). d The corresponding coplanar point set does not contain any points suitable for fitting, and is therefore irrelevant data; if the minimum Euclidean distance is Min... d If the distance is less than the distance threshold P, then the minimum Euclidean distance, Min, is used. d The two corresponding breakpoints are stored as one element in the filter point group.

[0078] Specifically, if point[n]∈counts[0] and point[n+1]∈counts[0], then point[n] and point[n+1] are coplanar. This process is repeated for all breakpoints, and the coplanar points form a coplanar point group: group[[point[1],point[2],...],...,[[point[n],point[n+1]]]. Then, all points within the coplanar point group group[n] are compared with all points in other point groups using a traversal Euclidean distance d. Finally, the minimum distance Min generated by the distance calculation between group[n] and other point groups is calculated. d According to the preset distance threshold P, if Min d If the distance is greater than the distance threshold P, then this point group has no other points to fit with. If Min d If the distance is less than P, and it is the minimum distance calculated between point group group[n] and points within other point groups, then the Min value is stored. d Points [u] and [i] are breakpoints. After traversing all point groups, a new point group new_group[[point[1],point[2],...],...,[[point[n],point[n+1]]] is formed by the points to be matched.

[0079] In a preferred embodiment, step S5 includes the following process for each element in the screening point group:

[0080] Using the two breakpoints stored in the element as center points, perform eight-neighbor analysis: if the sum of the eight neighboring pixel values ​​is equal to 1, then store the coordinates [k, l] of the points whose adjacent pixel values ​​are 1 into the set of points to be fitted (e.g., ...). Figure 8 (as shown);

[0081] Eight-neighbor analysis is performed with the point coordinates [k,l] as the center point: if the sum of the eight neighboring pixels of the point coordinates [k,l] is 2 (including the pixel value of the previous point), then the adjacent unstored point coordinates [k+1,l] of the point coordinates [k,l] are stored in the set of points to be fitted; if the sum of the eight neighboring pixels of the point coordinates [k,l] is greater than 2, then it is determined that there are redundant pixels interfering with the point coordinates [k,l].

[0082] As a preferred embodiment, please refer to Figure 9 In step S6, multivariate function fitting is performed according to the following formula:

[0083] F(x)=Ax 3 +Bx 2 +Cx+D.

[0084] Finally, the corresponding curves plotted in the predicted image and compared with the original remote sensing image can be found in [reference needed]. Figure 10 As can be seen, this embodiment can effectively handle and repair problems such as road discontinuity, noisy isolated patches, and layer holes caused by ground shadow interference from buildings, clouds, and trees, or layer breaks caused by trees covering the original road at both ends of the road, during the extraction of road feature information. It can draw road shapes that are close to reality and complete the drawing of complete road information.

[0085] Example 2

[0086] Please see Figure 11 A post-processing system for road extraction from remote sensing images includes, in sequence: a predicted image acquisition module 1, a skeleton line acquisition module 2, a breakpoint extraction module 3, an irrelevant data filtering module 4, a point set acquisition module 5, and a fitting and plotting module 6; wherein:

[0087] The predictive image acquisition module 1 is used to acquire remote sensing images, divide the remote sensing images into frames according to a preset cropping overlap, and perform road segmentation on the frame division results to obtain a predictive image containing roads and background.

[0088] The skeleton line acquisition module 2 is used to refine the predicted image and retain skeleton lines containing complete road structure information.

[0089] The breakpoint extraction module 3 is used to perform eight-neighbor analysis on the points on the skeleton line to extract the breakpoints on the skeleton line.

[0090] The irrelevant data filtering module 4 is used to filter out irrelevant data in the breakpoint by performing connected component analysis on the breakpoint.

[0091] The point set acquisition module 5 is used to perform eight-neighbor analysis on the filtering results of the irrelevant data filtering module 4, and exclude the breakpoints that are interfered by redundant pixels to form a point set to be fitted.

[0092] The fitting and drawing module 6 is used to fit the set of fitting points and draw the corresponding curves in the predicted image based on the fitting results to form complete road information.

[0093] Compared with existing technologies, the post-processing system for road extraction from remote sensing images of the present invention incorporates the post-processing of road extraction results into the feature extraction process, which to a certain extent improves the connectivity and integrity of road information, preserves and restores the linear information of roads to the maximum extent, and realizes the complete extraction of road information from remote sensing images.

[0094] Example 3

[0095] A storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the post-processing method for road extraction from remote sensing images as described in Example 1.

[0096] Example 4

[0097] A computer device includes a storage medium, a processor, and a computer program stored in the storage medium and executable by the processor, wherein the computer program, when executed by the processor, implements the steps of the post-processing method for road extraction from remote sensing images as described in Embodiment 1.

[0098] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A post-processing method for road extraction from remote sensing images, characterized in that, Includes the following steps: S1, acquire remote sensing images, divide the remote sensing images into sections according to a preset cropping overlap; perform road segmentation on the sectioning results to obtain a predicted image containing roads and background; S2, refine the predicted image, retaining the skeleton lines containing complete road structure information; The predicted image is first repaired using the following process, and then refined: After the predicted image undergoes grayscale, binarization, and dilation / erosion preprocessing, the independent patches in the predicted image are searched, and the area of ​​each independent patch is calculated. Independent patches with an area smaller than a preset area threshold are removed from the predicted image. Holes in the predicted image layer are filled. S3, perform eight-neighborhood analysis on the points on the skeleton line to extract the breakpoints on the skeleton line. S4. By performing connected component analysis on the breakpoint, irrelevant data in the breakpoint is filtered out. Specifically, connected component analysis is performed on the breakpoints to divide breakpoints belonging to the same independent graph tile into coplanar point groups; the minimum Euclidean distance between each coplanar point group and breakpoints outside the coplanar point group is obtained by calculating the Euclidean distance between each coplanar point group. d Iterate through the minimum Euclidean distance Min. d : If the minimum value of the Euclidean distance is Min d If the distance is greater than the preset distance threshold P, then the Euclidean distance is considered to be minimum (Min). d The corresponding coplanar point set does not contain any points suitable for fitting, and is therefore irrelevant data; if the minimum Euclidean distance is Min... d If the distance is less than the distance threshold P, then the minimum Euclidean distance, Min, is used. d The two corresponding breakpoints are stored as one element in the filter point group; S5, perform eight-neighborhood analysis on the filtering results of step S4 to exclude breakpoints that are interfered with by redundant pixels and form a set of points to be fitted. For each element in the filter point group, the following process is included: The two breakpoints stored in the element are used as center points for eight-neighbor analysis: if the sum of the pixel values ​​of the eight neighbors is equal to 1, the coordinates [k,l] of the points with adjacent pixel values ​​of 1 are stored in the set of points to be fitted. Eight-neighbor analysis is performed with the point coordinates [k,l] as the center point: if the sum of the eight neighboring pixels of the point coordinates [k,l] is 2, then the adjacent unstored point coordinates [k+1,l] of the point coordinates [k,l] are stored in the set of points to be fitted; if the sum of the eight neighboring pixels of the point coordinates [k,l] is greater than 2, then it is determined that there are redundant pixels interfering with the point coordinates [k,l]. S6, Fit the set of fitted points, and draw the corresponding curve in the predicted image based on the fitting result to form complete road information.

2. The post-processing method for road extraction from remote sensing images according to claim 1, characterized in that, In step S3, Includes the following processes: Calculate the sum of pixels in eight directions around a point on the skeleton line, and take the point where the sum of pixels is 1 as the break point on the skeleton line.

3. The post-processing method for road extraction from remote sensing images according to claim 1, characterized in that, In step S6, multivariate function fitting is performed according to the following formula: F(x)=Ax 3 +Bx 2 +Cx+D。 4. The post-processing method for road extraction from remote sensing images according to any one of claims 1 to 3, characterized in that, The cropping overlap is 100 pixels; the area threshold is 70*70.

5. A post-processing system for road extraction from remote sensing images, characterized in that, The module includes a sequentially connected predictive image acquisition module (1), a skeleton line acquisition module (2), a breakpoint extraction module (3), an irrelevant data filtering module (4), a point set acquisition module (5), and a fitting and plotting module (6); wherein: The predictive image acquisition module (1) is used to acquire remote sensing images, divide the remote sensing images into sections according to a preset cropping overlap, and perform road segmentation on the sectioning results to obtain a predictive image containing roads and background. The skeleton line acquisition module (2) is used to refine the predicted image and retain the skeleton line containing complete road structure information; The predicted image is first repaired using the following process, and then refined: After the predicted image undergoes grayscale, binarization, and dilation / erosion preprocessing, the independent patches in the predicted image are searched, and the area of ​​each independent patch is calculated. Independent patches with an area smaller than a preset area threshold are removed from the predicted image. Holes in the predicted image layer are filled. The breakpoint extraction module (3) is used to perform eight-neighbor analysis on the points on the skeleton line to extract the breakpoints on the skeleton line. The irrelevant data filtering module (4) is used to filter out irrelevant data in the breakpoint by performing connected component analysis on the breakpoint; Specifically, connected component analysis is performed on the breakpoints to divide breakpoints belonging to the same independent graph tile into coplanar point groups; the minimum Euclidean distance between each coplanar point group and breakpoints outside the coplanar point group is obtained by calculating the Euclidean distance between each coplanar point group. d Iterate through the minimum Euclidean distance Min. d : If the minimum value of the Euclidean distance is Min d If the distance is greater than the preset distance threshold P, then the Euclidean distance is considered to be minimum (Min). d The corresponding coplanar point set does not contain any points suitable for fitting, and is therefore irrelevant data; if the minimum Euclidean distance is Min... d If the distance is less than the distance threshold P, then the minimum Euclidean distance, Min, is used. d The two corresponding breakpoints are stored as one element in the filter point group; The point set acquisition module (5) is used to perform eight-neighbor analysis on the filtering results of the irrelevant data filtering module (4), and to exclude the breakpoints that are interfered by redundant pixels to form a point set to be fitted. For each element in the filter point group, the following process is included: The two breakpoints stored in the element are used as center points for eight-neighbor analysis: if the sum of the pixel values ​​of the eight neighbors is equal to 1, the coordinates [k,l] of the points with adjacent pixel values ​​of 1 are stored in the set of points to be fitted. Eight-neighbor analysis is performed with the point coordinates [k,l] as the center point: if the sum of the eight neighboring pixels of the point coordinates [k,l] is 2, then the adjacent unstored point coordinates [k+1,l] of the point coordinates [k,l] are stored in the set of points to be fitted; if the sum of the eight neighboring pixels of the point coordinates [k,l] is greater than 2, then it is determined that there are redundant pixels interfering with the point coordinates [k,l]. The fitting and drawing module (6) is used to fit the set of fitting points and draw the corresponding curves in the predicted image based on the fitting results to form complete road information.

6. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the post-processing method for road extraction from remote sensing images as described in any one of claims 1 to 4.

7. A computer device, characterized in that: The method includes a storage medium, a processor, and a computer program stored in the storage medium and executable by the processor, wherein the computer program, when executed by the processor, implements the steps of the post-processing method for road extraction from remote sensing images as described in any one of claims 1 to 4.

Images