A complex map segmentation method and system based on polygon segmentation

Abstract

The application provides a complex map segmentation method and system based on polygon segmentation, which comprises the following steps: performing image preprocessing on an input map to obtain a first map; extracting feature information of the first map, including contour and contour corner feature information of the first map; performing initial segmentation on the first map according to the number of contours contained in the first map, and adaptively adjusting and optimizing the initial segmentation effect to initially segment the first map into a plurality of map small blocks; fusing the map small blocks into map subblocks with uniform area and regular shape according to pre-input map subblock area and block number requirements, and labeling each map subblock in adjacent order to obtain a final segmentation result.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a method and system for segmenting complex maps based on polygon segmentation. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Map segmentation algorithms have significant application value in fields such as robot autonomous navigation, intelligent transportation, and environmental monitoring. It is a key technology for dividing large maps or environments into smaller, more manageable sub-maps or regions. Large maps contain complex structures and obstacles, making direct navigation and path planning on the entire map computationally intensive and time-consuming. Map segmentation divides the map into multiple smaller blocks, each with independent navigation and path planning, thereby improving navigation efficiency.

[0004] However, some existing map segmentation algorithms still face several problems. First, they are significantly affected by irregular map boundaries, leading to inaccurate segmentation results. Second, they are greatly affected by numerous obstacles, making it difficult to accurately divide regions. Furthermore, issues such as uneven segmentation areas, irregular segmentation shapes, and aesthetically unappealing results also exist. In complex structural environments, they can easily encounter situations where segmentation becomes unsolvable. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention provides a complex map segmentation method based on polygon segmentation, which can reduce the impact of irregular map boundaries on segmentation and is applicable to more diverse and complex unstructured terrain.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] Firstly, a method for segmenting complex maps based on polygon segmentation is disclosed, including:

[0008] The input map is preprocessed to obtain the first map;

[0009] Extract feature information from the first map, including the outline and corner feature information of the first map;

[0010] Based on the number of outlines contained in the first map, the first map is initially segmented, and the initial segmentation effect is adaptively adjusted and optimized to initially segment the first map into several small map pieces;

[0011] Based on the pre-input requirements for the area and number of map tiles, small map tiles are merged into map tiles with uniform area and regular shape, and each map tile is numbered in adjacent order to obtain the final segmentation result.

[0012] As a further technical solution, image preprocessing is performed on the input map, including:

[0013] Preprocess the input map by performing opening and then closing operations first;

[0014] Image opening and closing operations are composed of the combination of image dilation and erosion operations and set operations.

[0015] As a further technical solution, feature information of the first map is extracted, including:

[0016] Detect and extract all contours from the first map;

[0017] When the first map is a simple structure map, the number of outlines is 1, that is, there is only one outermost outline;

[0018] When the first map is a relatively complex map with multiple outlines, including an outermost main outline and several sub-outlines nested inside the main outline;

[0019] Corner points are points on the contour with a large rate of change; corner points are extracted from the contour.

[0020] The first map is initially segmented based on the extracted contours and corner points.

[0021] As a further technical solution, it also includes: adaptively adjusting the number and distribution of each contour corner point, specifically:

[0022] Given the desired number N of the main contour corner points after adjustment, delete the corner points on the main contour that are less than or equal to the set threshold D, and obtain a set of main contour corner points where the distance between any two adjacent corner points is greater than the threshold D and the total number of corner points meets the pre-given desired value N.

[0023] The corner points of each sub-contour are sparsified with a spacing D, so that the spacing between any two adjacent corner points on the sub-contour is also greater than D. The corner point spacing of the sub-contour is the same as that of the main contour, which is used to ensure that the number of corner points of contours with different perimeters is proportional after the sparsification adjustment.

[0024] As a further technical solution, it also includes: adaptively discarding unreasonable contours, the specific steps of which are as follows:

[0025] Step (1): Obtain all contours and extract the corner feature information of each contour, and perform equal spacing and sparsification processing on the corner points of each contour;

[0026] Given an expected value N for the number of corner points of a main contour, then sparsify each contour corner point to obtain a set of contour corner points whose number of corner points is proportional to the contour perimeter.

[0027] Step (2): Perform initial segmentation of the map according to each contour corner point. If there is no solution for segmentation, gradually increase N and repeat the above steps (1)-(2).

[0028] Step (3): If there is no solution for map segmentation until N is greater than the set value, then the smallest contour among all current contours is considered to be an unreasonable contour that causes no solution for segmentation, and it is discarded.

[0029] Step (4): After obtaining a new set of contours, N is reassigned to the expected value. Steps (1) to (3) are repeated for all contours. This process continues until a segmentation solution is found. At this point, all contours are reasonable contours that can be used for segmentation, and all unreasonable contours have been discarded.

[0030] As a further technical solution, the first map is initially segmented based on the number of contours it contains, including:

[0031] If the total number of contours is 1, i.e., in the case of a simple graph, then the optimal convex partitioning based on dynamic programming is adopted;

[0032] If the total number of contours is greater than 1, that is, in the case of a complex structure map where the main contour contains nested sub-contours, the Herttel-Melholm segmentation is used.

[0033] As a further technical solution, map tiles are merged into map sub-tiles with uniform area and regular shape, specifically:

[0034] Merge all map tiles with an area smaller than a set area threshold with their adjacent tiles;

[0035] The first map tile is selected based on the map tile index, and other map tiles are traversed. The other map tiles adjacent to the selected tile are merged into a map sub-tile. This process is repeated until the area of ​​the map sub-tile reaches the expected area size. Then the merging process of the next map sub-tile is started. That is, a new tile is selected from the remaining map tiles, and the above merging process is repeated.

[0036] At the same time, during this process, map sub-tiles are numbered according to their adjacent order;

[0037] After all map sub-tiles have been merged, all map sub-tiles are traversed again, and the map sub-tile with the smallest area is merged with its adjacent sub-tiles until the total number of map sub-tiles meets the expected number of tiles.

[0038] Secondly, a complex map segmentation system based on polygon segmentation is disclosed, including:

[0039] The map preprocessing and feature extraction module is configured to: perform image preprocessing on the input map using image processing technology, and extract the outline and outline corner feature information of the preprocessed map;

[0040] The input map segmentation and adaptive optimization module is configured to automatically select the corresponding segmentation algorithm to initially segment the map based on the number of contours contained in the map, and adaptively adjust and optimize the initial segmentation effect based on map features.

[0041] The map fusion module is configured to: merge small map tiles into uniformly sized, regularly shaped map tiles based on pre-inputted requirements for the area and number of map tiles, and label each map tile in adjacent order to obtain the final segmentation result.

[0042] The above one or more technical solutions have the following beneficial effects:

[0043] This invention discloses a complex map segmentation method based on polygon segmentation, which can reduce the impact of irregular map boundaries on segmentation, is applicable to more diverse and complex unstructured terrains, has strong robustness to obstacles and noise in the map, reduces the probability of segmentation failure due to complex map structure, reduces the problems of uneven map segmentation and irregular shape, can achieve automatic sequential labeling, improves practical application value, and does not require pre-training, making it a fast segmentation method.

[0044] This invention addresses the problems of irregular map boundaries and map noise by employing image processing techniques. It enhances the segmentation of complex and unstructured maps by using Hertel-Mehlhom segmentation and dynamic programming-based optimal convex segmentation. Furthermore, it improves the segmentation of complex and obstacle-laden maps by adaptively discarding excessively small map contours, reducing the occurrence of unsolvable segmentation problems. Finally, it effectively improves the regularity of initially segmented map patches by adaptively optimizing segmentation feature points. Finally, it integrates map patches into uniformly sized, regularly shaped sub-patterns and automatically numbers them according to their adjacent order.

[0045] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0046] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0047] Figure 1 This is a system framework diagram of a complex map segmentation method based on polygon segmentation according to an embodiment of the present invention;

[0048] Figure 2 This is an example of map preprocessing opening operation in an embodiment of the present invention. (a) is the original map, and (b) is the result of the opening operation.

[0049] Figure 3 This is a map preprocessing closing operation according to an embodiment of the present invention. (a) is the original map, and (b) is the result of the closing operation.

[0050] Figure 4(a) is a single-contour map of an embodiment of the present invention, and Figure 4(b) is a multi-contour map;

[0051] Figure 5(a) shows the extraction of corner points on the contour line of a single contour map, and Figure 5(b) shows the extraction of corner points on the contour line of a multi-contour map.

[0052] Figure 6(a) shows the corner point before adaptive optimization, and Figure 6(b) shows the corner point after adaptive optimization.

[0053] Figure 7(a) shows all initial contours, and Figure 7(b) shows contours after adaptively discarding excessively small contours.

[0054] Figure 8(a) shows the first type of division, and Figure 8(b) shows the first type of division;

[0055] Figure 9(a) Initial segmentation result of single contour map, Figure 9(b) Initial segmentation result of multi-contour map;

[0056] Figure 10(a) shows the initial segmentation result of the single contour map, and Figure 10(b) shows the fusion of small map patches.

[0057] Figure 11(a) shows the initial segmentation result of the multi-contour map, and Figure 11(b) shows the fusion of small map patches.

[0058] Figure 12(a) Initial segmentation result of single contour map, Figure 12(b) Result after fusion of small blocks, Figure 12(c) Final result;

[0059] Figure 13(a) Initial segmentation result of multi-contour map, Figure 13(b) Result after fusion of small blocks, Figure 13(c) Final result. Detailed Implementation

[0060] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0061] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0062] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0063] Example 1

[0064] See appendix Figure 1 As shown, this embodiment discloses a complex map segmentation method based on polygon segmentation, including:

[0065] Map preprocessing and feature extraction: Image processing techniques are used to perform operations such as image dilation and erosion, opening and closing operations on the input map to reduce the influence of map boundary irregularities and obstacles and noise in the map. Then, the contour and contour corner points of the optimized map are extracted.

[0066] Map segmentation and adaptive optimization: The system will automatically select the corresponding segmentation algorithm to perform initial segmentation of the map based on the number of contours contained in the map, and adaptively adjust and optimize the initial segmentation effect according to map features, reducing the occurrence of unsolvable segmentation due to complex map structures and large obstacles in the map;

[0067] Map fusion: Based on the pre-input map tile area and tile number requirements, small map tiles are merged into map tiles with uniform area and regular shape, and each map tile is numbered in adjacent order to obtain the final segmentation result.

[0068] Map preprocessing algorithm:

[0069] This algorithm preprocesses the input original map by performing opening and closing operations. Opening involves erosion followed by dilation of the target image; closing involves dilation followed by erosion. Opening helps reduce map jagged edges and mitigates the impact of irregular map boundaries on subsequent segmentation. Closing helps reduce the impact of noise points and small obstacles on segmentation. Image opening and closing operations are composed of the combination of dilation and erosion operations and set operations (union, intersection, complement, etc.). Opening generally smooths object contours, breaks narrow necks, and eliminates fine protrusions. The original image and the result of the opening operation are shown in the image below. Figure 2 (a) in the middle Figure 2As shown in (b) of the image. Through the opening operation, invalid areas such as thin edges and rough edges caused by laser light passing through shop windows and narrow corridors during robot mapping are disconnected from the main map content and are effectively eliminated, facilitating the extraction of more regular map contours. Next is the closing operation, which also smooths a portion of the contour. However, unlike the opening operation, it typically bridges narrow discontinuities and long, thin grooves, eliminates small holes, and fills in breaks in the image. The original image and the result of the closing operation are shown below. Figure 3 (a) in the middle Figure 3 As shown in (b) in the figure. Through the closing operation, the small black dots left by dynamic pedestrians and noise in the mall are effectively removed during the mapping process, thus reducing the impact of the dynamic environment and too many small obstacles in the mall on map segmentation.

[0070] Map feature extraction:

[0071] Feature extraction algorithms are used to extract features such as contours and corners from the preprocessed map. These features will be used as input for subsequent map segmentation. First, the map is converted into a binary image, and then the `cv2.findContours` function is used to retrieve contours from the binary image. After all contours of the map have been detected and extracted, a single-contour map is shown in Figure 4(a), and a multi-contour map is shown in Figure 4(b). When the map has a simple structure, the number of contours is 1, that is, there is only one outermost contour, as shown in Figure 4(a). For maps with more complex structures, there are usually multiple contours, typically including an outermost main contour and several sub-contours nested inside the main contour, as shown in Figure 4(b). The subsequent map segmentation module will select different segmentation algorithms to segment the map based on the total number of contours. Corners are points with a large rate of change on the contours. The `cv2.goodFeaturesToTrack` function can be used, which extracts high-quality corners from the image using the Shi-Tomasi method. The corners extracted from the contours are shown below. Figure 5(a) and 5(b) As shown. After the contours and corners are extracted, these map features will be input into the map segmentation module for the initial segmentation of the map.

[0072] Adaptive optimization algorithm: The adaptive optimization algorithm mainly has two functions: adaptively adjusting the number of corner points and adaptively discarding overly small contours.

[0073] The adaptive adjustment of corner point count is necessary because the map segmentation module initially segments the map using corner points. However, the original corner point distribution of each contour input by the feature extraction module is relatively dense and highly uneven, negatively impacting the aesthetics and uniformity of the final segmentation. Furthermore, the number of corner points for contours of different perimeters should be in a certain proportion; otherwise, irregular shapes of the segmented patches will result.

[0074] To address the aforementioned problems, this invention proposes an algorithm for adaptively adjusting the number and distribution of corner points in each contour: First, given the desired adjusted number N of corner points in the main contour, corner points on the main contour with a distance of less than or equal to D are deleted, resulting in a set of main contour corner points where the distance between any two adjacent corner points is greater than a threshold D, and the total number of corner points conforms to the pre-given desired value N. Next, using the same distance D, the corner points of each sub-contour are sparsified, ensuring that the distance between any two adjacent corner points on the sub-contour is also greater than D. The result before optimization is shown in Figure 6(a), and the result after optimization is shown in Figure 6(b). The corner point spacing of the sub-contour is the same as that of the main contour, ensuring that the number of corner points in contours with different perimeters is proportional after sparsification adjustment, which helps solve the problem of irregular shapes in corner-based segmentation blocks.

[0075] The adaptive discarding of excessively small outlines is because a map with a relatively complex structure may contain multiple large obstacles or multiple hollow structures. These factors result in too many small outlines inside the map, leading to severe unstructured phenomena and making it easy to encounter unsolvable segmentation problems.

[0076] To reduce the occurrence of unsolvable segmentation and enhance the robustness and stability of the algorithm for segmenting complex structured maps, this invention designs an algorithm that adaptively discards unreasonable contours.

[0077] First, this example obtains all contours, as shown in Figure 7(a). Then, the aforementioned adaptive corner point adjustment module performs equal spacing and sparsification processing on the corner points of each contour. Specifically, a desired number of corner points for the main contour, N = 150, is given, and then each contour corner point is sparsified, resulting in a set of contour corner points whose number is proportional to the contour perimeter. Then, the map is initially segmented based on each contour corner point. If the segmentation has a solution, the map fusion process proceeds. If no solution is found, N is gradually increased, and the above process is repeated.

[0078] If no solution for map segmentation is found until N > 1000, the smallest contour among all current contours can be considered an unreasonable contour causing the segmentation to fail. This smallest contour is discarded, and the remaining contours form a new set. After obtaining the new set of contours, N = 150 is reset, and the above process is repeated for all contours. This process continues until a segmentation solution is found. At this point, all contours are reasonable contours that can be used for segmentation, as shown in Figure 7(b), and all unreasonable contours have been discarded.

[0079] Map segmentation principle: After obtaining reasonable contours and adaptively optimized corner points for each contour, this data is input into the map segmentation algorithm to initially divide the entire map into several small map tiles. If the total number of contours is 1, i.e., a simple map, optimal convex segmentation based on dynamic programming is used. If the total number of contours is greater than 1, i.e., a complex map structure where the main contour contains nested sub-contours, this example uses Hertel-Mehlhom segmentation.

[0080] Optimal convex partitioning based on dynamic programming refers to partitioning a convex polygon into a set T of complementary intersecting triangles. The optimal convex partitioning is defined as follows: Given a convex polygon P and weight functions w defined on the triangles formed by the polygon's edges and chords, the goal is to determine the triangulation of the convex polygon such that the sum of the weights on all triangles in the triangulation is minimized. As shown in Figures 8(a) and 8(b), the first partitioning, starting from V3, has {v1v3, v3v0, v3v6, v4v6}; the partitioning starting from V1 has {v1v3, v1v4, v4v6, v1v6}. These are two forms of partitioning. The optimal partitioning is to find the partition where the sum of the edge weights w is minimized. In practical implementation, this dynamic programming-based optimal partitioning algorithm package is directly called for partitioning.

[0081] The Hertel-Mehlhorn algorithm is a polygon segmentation algorithm that divides a simple polygon into several non-intersecting convex polygons. Its core idea is to progressively add diagonals to subdivide a convex polygon into smaller convex polygons until no more diagonals can be added. Specifically, the algorithm first connects any two non-adjacent vertices of the polygon to form a diagonal, dividing the polygon into two sub-polygons. Then, the same segmentation process is recursively applied to each sub-polygon until no more diagonals can be added. This progressive segmentation method ensures that the segmented sub-polygons are all convex polygons. In practice, the Hertel-Mehlhorn algorithm package is directly called for segmentation.

[0082] The two algorithms described above initially divide the map into several small map patches, as shown in Figure 9(a) and Figure 9(b).

[0083] Map fusion: After obtaining several initially segmented map tiles, they need to be merged into map sub-tiles that meet the expected area size or expected total number of tiles. After the segmentation algorithm segments the map, due to the influence of complex terrain, the initial segmentation easily produces some excessively small map tiles. These excessively small map tiles will affect the efficiency and shape regularity of subsequent fusion into map sub-tiles. First, all excessively small map tiles are merged with adjacent tiles, such as... Figures 10(a)-10(b) and Figures 11(a)-11(b) As shown, this is to eliminate map tiles that are too small.

[0084] Then, the first map tile is selected based on its index, and other map tiles are traversed. Adjacent map tiles are merged with the selected tile to form a sub-tile. This process is repeated until the sub-tile's area reaches the expected size. The merging process for the next sub-tile is then initiated, selecting a new tile from the remaining tiles and repeating the process. Simultaneously, the sub-tiles are numbered according to their adjacent order. Because the indices of adjacent tiles are not necessarily continuous, and because the indices are interrupted at map edges, the resulting map tiles may have areas smaller than expected, leading to a total number of sub-tiles exceeding the expected size. This is a common problem in map segmentation: uneven segmentation and inaccurate tile counts. To address this, after all sub-tiles are merged, all tiles are traversed again, and the smallest sub-tile is merged with its adjacent tiles until the total number of sub-tiles meets the expected requirement. This effectively solves the problems of uneven segmentation and inaccurate tile counts. The final map segmentation result after map fusion module is as follows: Figures 12(a)-12(c) and Figures 13(a)-13(c) As shown.

[0085] Example 2

[0086] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0087] Example 3

[0088] The purpose of this embodiment is to provide a computer-readable storage medium.

[0089] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0090] Example 4

[0091] like Figure 1 As shown, the purpose of this embodiment is to provide a complex map segmentation based on polygon segmentation. The entire system can be divided into:

[0092] The system includes a map preprocessing and feature extraction module, a map segmentation and adaptive optimization module, and a map fusion module.

[0093] The map preprocessing and feature extraction module uses image processing techniques to perform image dilation, erosion, opening and closing operations on the input map to reduce the influence of map boundary irregularities and obstacles and noise in the map. Then, it extracts the contour and contour corners of the optimized map and inputs them into the map segmentation and adaptive optimization module.

[0094] The map segmentation and adaptive optimization module will automatically select the corresponding segmentation algorithm to initially segment the map based on the number of contours contained in the map, and adaptively adjust and optimize the initial segmentation effect according to map features, reducing the occurrence of unsolvable segmentation due to complex map structures and large obstacles in the map.

[0095] After the map is initially divided into several small map tiles, it is input into the map fusion module. Based on the pre-input requirements for the area and number of map tiles, the small map tiles are merged into map tiles with uniform area and regular shape. Each map tile is then labeled in adjacent order to obtain the final segmentation result.

[0096] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0097] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0098] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A complex map segmentation method based on polygon segmentation, characterized in that, include: The input map is preprocessed to obtain the first map; Extract feature information from the first map, including the outline and corner feature information of the first map; Extract feature information from the first map, including: Detect and extract all contours from the first map; When the first map is a simple structure map, the number of outlines is 1, that is, there is only one outermost main outline; When the first map is a relatively complex map with multiple outlines, including an outermost main outline and several sub-outlines nested inside the main outline; Corner points are points on the contour with a large rate of change; corner points are extracted from the contour. Based on the number of outlines contained in the first map, the first map is initially segmented, and the initial segmentation effect is adaptively adjusted and optimized to initially segment the first map into several small map pieces; Based on the pre-input requirements for the area and number of map tiles, small map tiles are merged into map tiles with uniform area and regular shape, and each map tile is numbered in adjacent order to obtain the final segmentation result. The adaptive adjustment and optimization of the initial segmentation effect includes: Step (1): Obtain all contours and extract the corner feature information of each contour, and perform equal spacing and sparsification processing on the corner points of each contour: Given the desired number of principal contour corner points after adjustment Based on this, delete items on the main contour that are less than or equal to a set threshold. The corner points are obtained by finding a distance greater than a threshold between any two adjacent corner points. The total number of corner points meets the pre-defined expected value. The set of main contour corner points; With spacing The corner points of each sub-contour are thinned out so that the distance between any two adjacent corner points on the sub-contour is also greater than 1. The corner spacing of the sub-contour is the same as that of the main contour. This is to ensure that the number of corner points of contours with different perimeters after sparsification adjustment is proportional, and finally obtain the corner point set of each contour whose number of corner points is proportional to the contour perimeter. Step (2): Perform initial segmentation of the map based on each contour corner point. If a segmentation solution is unavailable, then... Gradually increase, and repeat the above steps (1) - (2) again; Step (3): If until If the value is greater than the set value and there is no solution for map segmentation, then the smallest contour among all current contours is considered an unreasonable contour that causes no solution for segmentation and is discarded. Step (4): After obtaining a new set of contours, N is reassigned to the expected value. Steps (1) to (3) are repeated for all contours. This process continues until a segmentation solution is found. At this point, all contours are reasonable contours that can be used for segmentation, and all unreasonable contours have been discarded.

2. The complex map segmentation method based on polygon segmentation as described in claim 1, characterized in that, Image preprocessing of the input map includes: Preprocess the input map by performing opening and then closing operations first; Image opening and closing operations are composed of the combination of image dilation and erosion operations and set operations.

3. The complex map segmentation method based on polygon segmentation as described in claim 1, characterized in that, Based on the number of contours contained in the first map, the first map is initially segmented, including: If the total number of contours is 1, i.e., in the case of a simple graph, then the optimal convex partitioning based on dynamic programming is adopted; If the total number of contours is greater than 1, that is, in the case of a complex structure map where the main contour contains nested sub-contours, the Herttel-Melholm segmentation is used.

4. The complex map segmentation method based on polygon segmentation as described in claim 1, characterized in that, Small map tiles are merged into uniformly sized, regularly shaped map sub-tiles, specifically: Merge all map tiles with an area smaller than a set area threshold with their adjacent tiles; The first map tile is selected based on the map tile index, and other map tiles are traversed. The other map tiles adjacent to the selected tile are merged into a map sub-tile. This process is repeated until the area of ​​the map sub-tile reaches the expected area size. Then the merging process of the next map sub-tile is started. That is, a new tile is selected from the remaining map tiles, and the above merging process is repeated. At the same time, during this process, map sub-tiles are numbered according to their adjacent order; After all map sub-tiles have been merged, all map sub-tiles are traversed again, and the map sub-tile with the smallest area is merged with its adjacent sub-tiles until the total number of map sub-tiles meets the expected number of tiles.

5. A complex map segmentation system based on polygon segmentation, characterized in that, The complex map segmentation method based on polygon segmentation as described in any one of claims 1-4 includes: The map preprocessing and feature extraction module is configured to: perform image preprocessing on the input map using image processing technology, and extract the outline and outline corner feature information of the preprocessed map; The input map segmentation and adaptive optimization module is configured to automatically select the corresponding segmentation algorithm to initially segment the map based on the number of contours contained in the map, and adaptively adjust and optimize the initial segmentation effect based on map features. The map fusion module is configured to: merge small map tiles into uniformly sized, regularly shaped map tiles based on pre-inputted requirements for the area and number of map tiles, and label each map tile in adjacent order to obtain the final segmentation result.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method described in any of claims 1-4 above.

Images