A map matching method, device and storage medium

Abstract

Embodiments of the present specification provide a map matching method and device and a storage medium. The method comprises: acquiring a to-be-matched map; inputting the to-be-matched map into an image segmentation model to obtain a plurality of image segmentation mask graphs; determining, from the plurality of image segmentation mask graphs, a plurality of target mask graphs greater than a threshold value; performing positive and negative binaryzation processing on each target mask graph to obtain a positive and negative binaryzation mask graph; for each standard map in a preset standard map library, acquiring a standard map mask graph and a size of the standard map mask graph; generating a new positive and negative binaryzation mask graph of each target mask graph according to the standard map mask graph and the size; matching the generated new positive and negative binaryzation mask graph with the standard map mask graph based on an image matching model to respectively output a point pair of each target mask graph and the standard map mask graph; and determining a standard map matched with the to-be-matched map according to the most point pairs. The present application can realize intelligent matching of maps.

Description

A map matching method, apparatus and storage medium Technical Field

[0001] This document relates to the field of image processing technology, and in particular to a map matching method, apparatus and storage medium. Background Technology

[0002] As an abstract, symbolic, and visual representation of geospatial information, maps are undeniably important. With the development of information technology, map use has permeated all aspects of daily life, providing a variety of convenient services from navigation and food delivery to address lookup.

[0003] Existing technologies mainly utilize deep learning to identify the map to be identified, determine its features, and find similar maps based on these features, thereby determining the geographical location information of the map to be identified.

[0004] However, compared to the recognition of ordinary images, maps contain many geographical details and texture features. Furthermore, different maps have different texture features, production standards, and levels of complexity, and they overlap with each other. When using deep learning to recognize maps, there is a lot of noise, the feature extraction accuracy is low, and it is difficult to quickly and accurately recommend standard map references for the map to be matched. Summary of the Invention

[0005] In view of the above solutions, this application aims to provide a map matching method, apparatus and storage medium to solve at least one of the above technical problems.

[0006] Firstly, one or more embodiments of this specification provide a map matching method, including:

[0007] S1. Obtain the map to be matched;

[0008] S2. Input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps;

[0009] S3. From the plurality of image segmentation masks, determine a plurality of target masks that are greater than a threshold;

[0010] S4. Perform positive and negative binarization processing on each of the target mask images to obtain a positive binarized mask image and a negative binarized mask image;

[0011] S5. For each standard map in the preset standard map library, obtain the standard map mask and the size of the standard map mask;

[0012] S6. For each standard map in the preset standard map library, generate a new positive binary mask and a new negative binary mask for each target mask based on the standard map mask and the size; wherein, the size represents the number of pixels in height and the number of pixels in width;

[0013] S7. Based on the image matching model, the generated new positive binary mask and the new inverse binary mask are matched with the standard map mask, respectively, and the first matching point pair and the second matching point pair of each target mask and the standard map mask are output respectively. The first matching point pair represents the matching point pair between the new positive binary mask of each target mask in the map to be matched and the standard map mask, and the second matching point pair represents the matching point pair between the new inverse binary mask of each target mask in the map to be matched and the standard map mask; and

[0014] S8. Based on the first matching point pair and the second matching point pair of each of the target mask maps, determine the standard map that matches the map to be matched.

[0015] Further, the step of inputting the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps includes:

[0016] The map to be matched is divided into grids to generate multiple prompt points;

[0017] Based on an image segmentation model, using multiple cue points, the image texture and tile features of the map to be matched are extracted; and

[0018] Based on the image texture and the tile features, the map to be matched is segmented into multiple image segmentation masks.

[0019] Further, determining multiple target masks greater than a threshold based on the multiple image segmentation mask images includes:

[0020] Determine the area of ​​each of the image segmentation masks;

[0021] The image segmentation mask is sorted according to the area; and

[0022] The image segmentation mask image with a value greater than a threshold is selected as the target mask image.

[0023] Furthermore, the step of generating a new positive binarized mask and a new inverse binarized mask for each target mask map based on the standard map mask map and the size for each standard map in the preset standard map library includes:

[0024] If the width and height in pixels of the standard map mask are different,

[0025] Based on the standard map mask, obtain the height in pixels and the width in pixels of the standard map mask;

[0026] Determine the minimum number of pixels between the height pixel count and the width pixel count in the standard map mask;

[0027] The number of pixels in each target mask image is obtained separately; wherein, the number of pixels in the target mask image includes the number of pixels in the height of the target mask image and the number of pixels in the width of the target mask image;

[0028] A first scaling reference is determined based on the minimum number of pixels in the standard map mask, wherein the first scaling reference is the height or width of the standard map mask;

[0029] The first adjustment ratio is determined based on the first ratio adjustment benchmark and the number of pixels in each of the target mask images;

[0030] According to the first adjustment ratio, the number of pixels in width and the number of pixels in height of the positive binarized mask and the inverse binarized mask of each target mask are enlarged or reduced proportionally to generate a new positive binarized mask and a new inverse binarized mask.

[0031] If the width and height pixels of the standard map mask are the same,

[0032] Based on the target mask image, the number of pixels in height and the number of pixels in width of the target mask image are obtained;

[0033] Determine the minimum pixel among the height pixel count and the width pixel count in the target mask image;

[0034] A second proportional adjustment reference is determined based on the minimum number of pixels in the target mask image, wherein the second proportional adjustment reference is the height or width in the target mask image;

[0035] The second adjustment ratio is determined based on the second ratio adjustment benchmark and the number of pixels in each of the standard map mask images;

[0036] According to the second adjustment ratio, the number of pixels in width and height of the positive binarized mask and the inverse binarized mask of each target mask are enlarged or reduced proportionally to generate a new positive binarized mask and a new inverse binarized mask.

[0037] Furthermore, the step of matching the generated new positive binary mask and the new inverse binary mask with the standard map mask based on the image matching model, and outputting the first and second matching point pairs of each target mask and the standard map mask respectively, includes: the image matching model includes a feature detector and a graph neural network model, and...

[0038] Using a feature detector, point and line segment features are extracted from the new positive binarized mask, the new inverse binarized mask, and the standard map mask; and

[0039] Based on the graph neural network model, the point and line segment features of the new positive binary mask and the new inverse binary mask are matched with the point and line segment features of the standard map mask, respectively, and the first matching point pair and the second matching point pair between the new positive binary mask and the new inverse binary mask and the standard map mask are output respectively.

[0040] Furthermore, the preset standard map library includes multiple standard maps; each standard map is binarized to obtain multiple standard map masks; and

[0041] The step of matching the generated new positive binary mask and the new inverse binary mask with the standard map mask based on the image matching model, and outputting the first matching point pair and the second matching point pair of each target mask and standard map mask respectively, includes:

[0042] For each standard map mask, determine the size of the current standard map mask;

[0043] Based on the current size of the standard map mask, update the sizes of the positive binarization mask and the inverse binarization mask to obtain the new positive binarization mask and the new inverse binarization mask;

[0044] Based on the current standard map mask and the new positive binary mask, the first matching point pair is determined; and

[0045] The second matching point pair is determined based on the current standard map mask and the new inverse binarized mask.

[0046] Further, determining the standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each of the target mask maps includes:

[0047] Sort the first and second matching point pairs of each of the standard map masks and each of the target masks respectively, and output the standard map with the most matching point pairs when matching the target mask.

[0048] Furthermore, steps S5 and S6 to S8 are executed using a distributed computing framework.

[0049] Secondly, embodiments of this application provide a map matching device, including:

[0050] The first acquisition module is used to acquire the map to be matched;

[0051] The first determining module is used to input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps;

[0052] The second determining module is used to determine multiple target masks with values ​​greater than a threshold from the multiple image segmentation mask images;

[0053] The third determining module is used to perform positive and negative binarization processing on each of the target mask images to obtain a positive binarized mask image and a negative binarized mask image;

[0054] The second acquisition module is used to acquire a standard map mask and the size of the standard map mask for each standard map in the preset standard map library;

[0055] The fourth determining module is used to generate new positive binary masks and new inverse binary masks for each of the target masks based on the standard map mask and the dimensions; wherein the dimensions represent the number of pixels in height and the number of pixels in width; and

[0056] The fifth determining module is used to match the generated new positive binary mask and the new inverse binary mask with the standard map mask based on an image matching model, and output a first matching point pair and a second matching point pair for each target mask and the standard map mask, wherein the first matching point pair represents the matching point pair between the new positive binary mask of each target mask in the map to be matched and the standard map mask, and the second matching point pair represents the matching point pair between the new inverse binary mask of each target mask in the map to be matched and the standard map mask; and

[0057] The sixth determining module is used to determine a standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each of the target mask maps.

[0058] Thirdly, embodiments of this application provide a storage medium for storing computer-executable instructions, characterized in that the computer-executable instructions, when executed, are the steps of the map matching method described in any one of the first aspects.

[0059] Compared with the prior art, this application can achieve at least the following technical effects:

[0060] This application utilizes an image segmentation model and a map to be matched to divide a map into multiple parts, quickly locating the target area of ​​interest to the user from these maps, thereby reducing noise during map use and improving search speed. Then, it employs positive and negative binarization techniques to process the image from different angles, highlighting different edges and textures to facilitate subsequent map matching. Next, it modifies the size of the positive and negative binarized masks to match the standard map mask, reducing matching errors during map matching. Finally, it uses an image matching model to output matching point pairs between the positive and negative binarized masks and the standard map mask, ensuring as many matching point pairs as possible through the aforementioned positive and negative binarization and size modification. Finally, it uses sorting to quickly find a standard map that matches the map to be matched, and further improves matching accuracy and efficiency through the aforementioned positive and negative binarization techniques and map size modification. In summary, by performing positive and negative binarization on the map to be processed and modifying its size, and comparing the modified map with a standard map, the most matching standard map can be accurately found from among the various standard maps. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 is a flowchart of a map matching method provided by one or more embodiments of this specification;

[0063] Figure 2 is a schematic diagram of the map segmentation process of a map matching method provided in one or more embodiments of this specification;

[0064] Figure 3 is a schematic diagram of the positive and negative binarization process of a map matching method provided in one or more embodiments of this specification;

[0065] Figure 4 is a schematic diagram of matching a positive binary mask image with a standard map mask image provided in one or more embodiments of this specification;

[0066] Figure 5 is a schematic diagram of the matching between the inverse binarized mask and the standard map mask provided in one or more embodiments of this specification;

[0067] Figure 6 is a schematic diagram of a map matching device provided in one or more embodiments of this specification. Detailed Implementation

[0068] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0069] Image matching aims to identify and align content or structures with identical or similar attributes in two images at the pixel level. In the current context of big data on the internet, image matching of maps can be used to identify areas represented by maps, compare differences in main details between map images, and thus uncover potentially problematic map images on the internet. This has significant meaning and role in the development of the geographic information industry and the regulation of internet maps. With the rapid development of the internet, various types of maps are disseminated online. Inconsistent maps can lead to cognitive errors, thereby affecting normal social and economic activities.

[0070] To address the technical problems existing in the prior art, this application proposes a map matching method, as shown in Figure 1, with the following specific steps:

[0071] Step S1: Obtain the map to be matched.

[0072] In this embodiment of the application, the system determines and obtains the map to be matched based on the map uploaded by the user.

[0073] Step S2: Input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps.

[0074] In this embodiment of the application, based on the image segmentation model (SAM, Segment Anything Model), texture features and tile features on the map image to be matched are extracted, and the map image to be matched is segmented to generate multiple image segmentation masks.

[0075] Specifically, the SAM image segmentation model is trained on a large dataset of 11 million images containing 1 billion masks. On unknown datasets and tasks, SAM exhibits excellent zero-trigger segmentation performance and can solve a wide range of segmentation tasks. Its segmentation process is driven by different cues, such as points or bounding boxes, to indicate the pixel semantics and region location of target objects. This application generates multiple cue points for the SAM model by uniformly gridding the map image and using the grid center points as cue points. These cue points are then input into the SAM model to segment the map image, generating image segmentation masks for multiple targets. Furthermore, the uniform gridding cue point generation strategy improves the map target segmentation coverage, as shown in Figure 2.

[0076] For example, the map image to be identified contains regions such as Beijing, Hebei, and Chongli. Each region is distinguished by lines. Using the SAM image segmentation model, the image is segmented according to the texture and tile features of the map to be matched, and boundary maps of the image segmentation mask images of Beijing, Hebei, Chongli, etc. are obtained respectively.

[0077] Step S3: From the plurality of image segmentation masks, determine a plurality of target masks that are greater than a threshold.

[0078] In this embodiment of the application, each image segmentation mask is traversed to obtain the size of each image segmentation mask and calculate the area of ​​each image segmentation mask. The image segmentation masks are sorted according to the area and the target mask with a value greater than the threshold is selected.

[0079] For example, given image segmentation masks: Image 1, Image 2, Image 3, and Image 4, each image segmentation mask is opened using an image processing library (such as PIL or OpenCV), and its dimensions (width and height) are read. Image 1 has a width of 300 pixels and a height of 200 pixels; Image 2 has a width of 800 pixels and a height of 600 pixels; Image 3 has a width of 1920 pixels and a height of 1080 pixels; and Image 4 has a width of 5000 pixels and a height of 3333 pixels. The area is then calculated based on the width and height of each image: Image 1 has an area of ​​60,000 pixels; Image 2 has an area of ​​480,000 pixels; Image 3 has an area of ​​2,073,600 pixels; and Image 4 has an area of ​​16,665,000 pixels. The images are sorted by area: Image 4 > Image 3 > Image 2 > Image 1. If the threshold is 48,000 pixels, then Images 4, 3, and 2 are determined to be possible target masks.

[0080] Step S4: Perform positive and negative binarization processing on each of the target mask images to obtain positive binarized mask images and negative binarized mask images.

[0081] In this embodiment of the application, the target object is binarized with 0 and 1, and the inside of the target object is assigned a value of 1 or the outside of the target object is assigned a value of 1. The binarization conversion is performed to generate two different binarized images, which are respectively obtained as positive binarization mask and inverse binarization mask.

[0082] Specifically, the target segmentation mask is first binarized, and then the binarized mask is debinarized. Furthermore, different map images have varying resolutions, resulting in differences in the details of the SAM model segmentation boundaries for the same region, as shown in Figure 3. For example, differences in the smoothness and curvature of the map target segmentation boundaries can significantly limit the matching effect of the map boundaries.

[0083] For example: First, find the contour of the target segmentation mask. By assigning a value of 1 (white) to the inside of the contour and a value of 0 (black) to the outside of the contour, a positive binary mask is obtained. Then, by assigning a value of 0 (black) to the inside of the contour and a value of 1 (white) to the outside of the contour, an inverse binary mask is obtained. The positive and inverse binary masks have different degrees of smoothness and curvature relative to the boundary, which will recommend different matching results, thereby improving the accuracy of map matching.

[0084] Step S5: For each standard map in the preset standard map library, obtain the standard map mask and the size of the standard map mask.

[0085] In this embodiment of the application, each standard map in the standard map library is traversed. The maps in the standard map library are also based on the image segmentation tool SAM. Each object of the standard map is segmented and its corresponding mask is obtained. The mask is then binarized to obtain the standard map mask. The pixel size of the standard map mask is then determined.

[0086] Specifically, the preset standard map library includes standard maps and standard map masks. The standard map masks are obtained by segmenting each standard map using the SAM image segmentation model to obtain multiple image segmentation masks. Then, the multiple image segmentation masks are binarized to obtain the standard map mask.

[0087] Step S6: Based on the standard map mask and the dimensions, generate a new positive binarized mask and a new inverse binarized mask for each of the target masks; wherein, the dimensions represent the number of height pixels and the number of width pixels.

[0088] In this embodiment, the pixel dimensions of each standard map mask, the pixel dimensions of the positive binarization mask, and the pixel dimensions of the inverse binarization mask are read. The pixel dimensions of the positive binarization mask and the inverse binarization mask are adjusted to match the pixel dimensions of the standard map mask while maintaining their aspect ratio. That is, based on the standard map mask (binarization mask) in the standard map library, the positive binarization mask and the inverse binarization mask to be matched are proportionally reduced or enlarged according to the minimum number of pixels in length and width of the standard map binarization mask to generate a multi-scale binarization mask. This ensures that the scale of the new positive binarization mask and the new inverse binarization mask to be matched is consistent with the standard map binarization mask, thus generating the final multi-target binarization mask to be matched.

[0089] For example, the preset standard map library includes standard map mask 1, standard map mask 2, standard map mask 3, etc. Standard map mask 1 has a width of 300 pixels and a height of 200 pixels; standard map mask 2 has a width of 400 pixels and a height of 300 pixels; standard map mask 3 has a width of 100 pixels and a height of 100 pixels.

[0090] The width and height of the positive and negative binarized masks are 100 pixels and 200 pixels respectively. The minimum pixel size (height or width) of the standard map masks 1, 2, and 3 is determined. The pixel sizes of the positive and negative binarized masks are adjusted according to the proportions corresponding to the minimum pixel size of each mask. The adjusted new positive and negative binarized masks have the same proportions as the standard map masks, with widths of 100 pixels and heights of 200 pixels; widths of 150 pixels and heights of 300 pixels; and widths of 100 pixels and heights of 200 pixels respectively. These are then used for subsequent recognition.

[0091] The calculation process is as follows: When the width and height of the standard map mask are the same, the width of the standard map mask 3 is 100 pixels and the height is 100 pixels; when the width of the positive and negative binarized mask is 100 pixels and the height is 200 pixels, the minimum number of pixels in the positive and negative binarized mask is determined to be the width. Based on the width and pixel count of the positive and negative binarized mask and the width and pixel count of the standard map mask, the adjustment ratio is determined to be 1. Therefore, the adjusted new positive and negative binarized mask has the same proportional size as the standard map mask, with a width of 150 pixels and a height of 300 pixels respectively.

[0092] When the width and height in pixels of the standard map mask are different, and the width of the standard map mask 2 is 300 pixels and the height is 200 pixels, while the width of the inverted binarized mask is 100 pixels and the height is 200 pixels, the minimum number of pixels in the standard map mask 1 is determined as the height. Based on the height in pixels of the standard map mask 1 and the height in pixels of the inverted binarized mask, the adjustment ratio is determined to remain unchanged at 1x. Therefore, the adjusted new inverted binarized mask has the same proportional size as the standard map mask, with a width of 100 pixels and a height of 200 pixels.

[0093] Step S7: Based on the image matching model, the generated new positive binarized mask and the new inverse binarized mask are matched with the standard map mask, respectively, and the first matching point pair and the second matching point pair of each target mask and the standard map mask are output respectively. The first matching point pair represents the matching point pair between the new positive binarized mask of each target mask in the map to be matched and the standard map mask, and the second matching point pair represents the matching point pair between the new inverse binarized mask of each target mask in the map to be matched and the standard map mask.

[0094] In this embodiment, the modified new positive binarized mask and the new inverse binarized mask are input together with the standard map mask into the image matching model (GlueSick). Points and line segments of the new positive binarized mask and the new inverse binarized mask and the standard map mask are extracted respectively. The first matching point pair of the new positive binarized mask and the standard map mask, and the second matching point pair of the new inverse binarized mask and the standard map mask are determined.

[0095] Specifically, common feature detectors are used to extract points, lines, and their descriptors, which are then uniformly encoded into a wireframe. A graph neural network is then used to encode the appearance and spatial information of the input features, fully exploring the relationships between points and outputting matching feature descriptors with more information. Finally, two independent Softmax layers are used to match points and lines separately, determining the matching correspondences. By processing two types of features together, the traditional paradigm of treating matching points and matching line segments between two views as two independent problems is broken. The correspondences between points and lines are jointly established within a single framework, resulting in a comprehensive improvement in the matching performance of the GlueStick method for points and lines. Based on the binarized mask of the multi-object to be matched, a cyclic traversal approach is used to match it with the binarized mask of the standard map. Based on a majority matching strategy, the matching pair with the most matching points is selected as the best match, recommending the most suitable standard map for the map to be matched.

[0096] Step S8: Determine a standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each target mask map.

[0097] In this embodiment, a first matching point pair and a second matching point pair are obtained for multiple standard map masks and the positive binarized mask and inverse binarized mask of each target mask, respectively. The first matching point pairs and the second matching point pairs obtained under multiple target masks are sorted in a unified manner to determine the standard map mask with the most matching point pairs. Then, the standard map that matches the map to be matched can be determined as the recommended map.

[0098] For example: In target mask image 1, the new positive binarized mask has 50 points and the new inverse binarized mask has 40 points; in target mask image 2, the new positive binarized mask has 60 points and the new inverse binarized mask has 35 points; standard map mask image 1 has 45 points and standard map mask image 2 has 50 points.

[0099] Case 1: Associating the new positive binary mask in target mask image 1 with standard map 1 can correspond to 40 points;

[0100] Scenario 2: Associating the new inverse binarized mask in target mask image 1 with standard map 1 will correspond to 30 points;

[0101] Case 3: Associating the new positive binary mask in target mask image 1 with standard map 2 can correspond to 45 points;

[0102] Case 4: Associating the new inverse binarized mask in target mask image 1 with standard map 2 will correspond to 20 points;

[0103] Case 5: Associating the new positive binary mask in target mask image 2 with standard map 1 will correspond to 15 points;

[0104] Case 6: Associate the new inverse binarized mask in target mask image 2 with standard map 1, which can correspond to 10 points;

[0105] Case 7: Associating the new positive binary mask in target mask image 2 with standard map 2 will correspond to 5 points;

[0106] Case 8: Associating the new inverse binarized mask in target mask image 2 with standard map 2 will correspond to 25 points;

[0107] Then, based on the matching point pairs, the order is: Case 3 > Case 1 > Case 2 > Case 8 > Case 4 > Case 5 > Case 6 > Case 7. Therefore, Case 3 has the most matching points and is the best match. The standard map in the standard map mask 2 is the standard map that best matches the map to be matched.

[0108] As shown in Figures 4 and 5, by performing forward and reverse binarization on the target mask image, the complex information in the image is simplified into black and white. This makes the foreground and background clearly separated during the comparison of point and line segment features, enhances the contrast of the image, makes the parts of interest more prominent, and can be matched with the standard map from different angles, thus improving the accuracy of finding the standard map.

[0109] Furthermore, the map to be matched is divided into grids to generate multiple cue points; based on the image segmentation model, the image texture and tile features of the map to be matched are extracted using the multiple cue points; and based on the image texture and the tile features, the map to be matched is segmented into multiple image segmentation masks.

[0110] Specifically, the map to be matched is divided into multiple small grids according to certain rules (such as equal spacing or adaptive based on the complexity of map content), and cue points are generated at the center or key locations of each grid. These cue points serve as reference points for subsequent image processing and feature extraction, helping to more accurately locate and identify key elements in the map. Using the SAM image segmentation model, different elements in the map (Beijing, Guangdong, Inner Mongolia, etc.) are identified and distinguished. Texture features of the grid regions are extracted by calculating texture descriptors such as the gray-level co-occurrence matrix (GLCM) and local binary mode (LBP) of pixels within the grid. Combined with the location of the cue points, a more refined analysis is performed on the area surrounding each grid or cue point, extracting tile features such as shape, size, and color. These features further describe the detailed attributes of elements in the map. The extracted image texture features and tile features are then fused to form a comprehensive feature representation of each grid or cue point region. Each grid or cue point region is then classified based on the fused features. Finally, based on the classification results, a category label is assigned to each pixel of the map to be matched, generating multiple image segmentation masks.

[0111] Further, the area of ​​each of the image segmentation masks is determined; the image segmentation masks are sorted according to the area; and the image segmentation masks with an area greater than a threshold are selected as target masks.

[0112] Specifically, for each image segmentation mask, the area is determined, and then a sorting algorithm is used to sort the image segmentation masks according to their area size. A threshold is used to select the masks with larger areas as the target masks.

[0113] Further, if the width and height pixel counts of the standard map mask are different, the height pixel count and width pixel count of the standard map mask are obtained based on the standard map mask; the minimum pixel count between the height and width pixel counts in the standard map mask is determined; the pixel count of each target mask is obtained respectively; wherein the pixel count of the target mask includes the height pixel count and the width pixel count of the target mask; a first proportional adjustment benchmark is determined based on the minimum pixel count in the standard map mask, the first proportional adjustment benchmark being the height or width of the standard map mask; a first adjustment ratio is determined based on the first proportional adjustment benchmark and the pixel count of each target mask; based on the first adjustment ratio, the width pixel count and height pixel count of the positive binarized mask and the inverse binarized mask of each target mask are adjusted respectively. The following steps are performed: 1. Scale up or down proportionally to generate new positive binary masks and new negative binary masks. 2. If the width and height pixels of the standard map mask are the same, obtain the height and width pixels of the target mask based on the target mask. 3. Determine the minimum pixel among the height and width pixels in the target mask. 4. Determine a second scaling reference based on the minimum pixel count in the target mask, where the second scaling reference is the height or width of the target mask. 5. Determine a second scaling ratio based on the second scaling reference and the pixel count of each standard map mask. 6. Scale up or down proportionally the width and height pixels of each target mask's positive and negative binary masks based on the second scaling ratio to generate new positive and negative binary masks.

[0114] Specifically, in scenario one, if the width and height pixel counts of the standard map mask differ, firstly, the height and width pixel counts are obtained from the standard map mask. Among these, the smaller value is selected as the minimum pixel count, which will serve as the benchmark for subsequent adjustments. Next, the height and width pixel counts of each target mask are obtained, and using the minimum pixel count as the benchmark, a scaling factor is calculated for each target mask relative to the standard map mask. Secondly, based on the calculated scaling factor, the width and height of the positive binarized mask and the inverse binarized mask for each target mask are proportionally enlarged or reduced, ensuring that the dimensions of all target masks maintain a certain proportional relationship with the standard map mask, thus making subsequent point pair association more accurate. Finally, after the above adjustments, each target mask will generate a new positive binarized mask and a new inverse binarized mask, and these new masks will have a similar proportional relationship to the standard map mask. Scenario 2: If the width and height pixels of the standard map mask are the same, compare the height and width pixels in the target mask to find the minimum number of pixels, then calculate the adjustment ratio to generate a new positive binarized mask and a new inverse binarized mask.

[0115] Furthermore, the image matching model includes a feature detector and a graph neural network model. The feature detector extracts point and line segment features from the new positive binary mask, the new inverse binary mask, and the standard map mask. Based on the graph neural network model, the point and line segment features of the new positive binary mask and the new inverse binary mask are matched with the point and line segment features of the standard map mask, respectively, and the first and second matching point pairs between the new positive binary mask and the new inverse binary mask and the standard map mask are output.

[0116] Specifically, a feature detector is used to first extract feature points and line segments from the new positive binary image and the new negative binary image. Then, point and line segment features are extracted from the standard map images in a pre-defined standard map library. A graph neural network model is then used to match the new positive and negative binary point and line segment features with the point and line segment features in the standard map library, thus obtaining the first matching point pairs between the new positive binary image and each standard map image in the pre-defined standard map library, and the second matching point pairs between the new negative binary image and each standard map image in the pre-defined standard map library. This method ensures that all standard maps in the pre-defined standard map library are matched with the map to be matched, thereby reducing the possibility of inaccurate matching results.

[0117] Furthermore, the preset standard map library includes multiple standard maps; each standard map is binarized to obtain multiple standard map masks; for each standard map mask, the size of the current standard map mask is determined; based on the size of the current standard map mask, the sizes of the positive binarized mask and the inverse binarized mask are updated to obtain the new positive binarized mask and the new inverse binarized mask; based on the current standard map mask and the new positive binarized mask, the first matching point pair is determined; and based on the current standard map mask and the new inverse binarized mask, the second matching point pair is determined.

[0118] Specifically, each standard map in each preset standard map library is binarized to obtain a standard map smear. The dimensions of each standard map smear, as well as the dimensions of the positive binarized smear and the negative binarized smear, are measured. Before comparing the test map with the standard map, the dimensions of the positive binarized smear are modified based on the dimensions of the standard map smear. Then, point and line segment features on the boundaries of the standard map smear and the new positive binarized smear are extracted. The point and line segment features on the standard map smear are paired with those on the new positive binarized smear to obtain multiple matching point pairs (first matching point pairs). Then, the dimensions of the negative binarized smear are modified based on the dimensions of the standard map smear. The point and line segment features on the boundaries of the standard map smear and the new negative binarized smear are extracted. The point and line segment features on the standard map smear are paired with those on the new negative binarized smear to obtain multiple matching point pairs (second matching point pairs).

[0119] Furthermore, the first and second matching point pairs of each of the standard map masks and each of the target masks are sorted, and the standard map with the most matching point pairs when matching the target mask is output.

[0120] Specifically, a pre-defined standard map library contains multiple standard map splices. Each standard map splice is matched against the new positive binarized splice and the new inverse binarized splice in the map to be matched, resulting in a first matching point pair and a second matching point pair. It should be noted that different standard map splices yield different first and second matching point pairs when matched against the new positive and inverse binarized splices. These different first and second matching point pairs are then sorted, and the one with the most matching point pairs is selected and its standard map is output. This standard map is the one most similar to the map to be matched.

[0121] Furthermore, steps S5 and S6 to S8 are performed using a distributed computing framework.

[0122] Specifically, after obtaining the standard map images from the preset standard map library, distributed computing is used to traverse the standard maps in the preset standard map library, and binarize each standard map in parallel to determine the size of the binarized standard map images. The sizes of the positive binarized images and the negative binarized images are modified according to the size. Point and line segment features of the standard map images, the new positive binarized images, and the new negative binarized images are extracted. The point and line segment features of the new positive binarized images and the new negative binarized images are matched with the point and line segment features of the standard map images to obtain different first matching point pairs and second matching point pairs. These matching point pairs are then sorted to obtain the standard map.

[0123] For example, a pre-defined standard map library contains Standard Map 1, Standard Map 2, and Standard Map 3. These are binarized to obtain Standard Map Mask 1, Standard Map Mask 2, and Standard Map Mask 3. During parallel processing, multiple nodes simultaneously perform matching operations between multiple new positive and negative binarized masks and the standard maps. The standard map binarized masks are pre-stored to reduce the need for multiple segmentation and masking processes during each matching process, thus improving efficiency. Specifically, node one modifies the dimensions of the positive and negative binarized masks to match those of Standard Map Mask 1. Then, it matches the point and line segment features of the new positive and negative binarized masks with the point and line segment features of Standard Map Mask 1 to obtain the first and second matching point pairs. Node 2: Modify the dimensions of the positive and negative binarized mask images to match those of the standard map mask image 2. Then, obtain the standard map mask image 2 again. Match the point and line segment features of the new positive and negative binarized mask images with the point and line segment features of the standard map mask image 2 to obtain the first and second matching point pairs. Node 3: Modify the dimensions of the positive and negative binarized mask images to match those of the standard map mask image 3. Then, obtain the standard map mask image 3 again. Match the point and line segment features of the new positive and negative binarized mask images with the point and line segment features of the standard map mask image 3 to obtain the first and second matching point pairs. The above matching is performed iteratively on the positive and negative binarized mask images of multiple targets. Then, the first and second matching point pairs for each channel are uniformly sorted to obtain the standard map most similar to the map to be matched.

[0124] This application provides a map matching device, as shown in Figure 6, including:

[0125] The first acquisition module 101 is used to acquire the map to be matched;

[0126] The first determining module 102 is used to input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps;

[0127] The second determining module 103 is used to determine multiple target masks with values ​​greater than a threshold from the multiple image segmentation mask images;

[0128] The third determining module 104 is used to perform positive and negative binarization processing on each of the target mask images to obtain a positive binarized mask image and a negative binarized mask image.

[0129] The second acquisition module 105 is used to acquire a standard map mask and the size of the standard map mask for each standard map in the preset standard map library.

[0130] The fourth determining module 106 is configured to generate new positive binary masks and new inverse binary masks for each of the target masks based on the standard map mask and the dimensions; wherein the dimensions represent the number of pixels in height and the number of pixels in width; and

[0131] The fifth determining module 107 is used to match the generated new positive binary mask and the new inverse binary mask with the standard map mask based on an image matching model, and output a first matching point pair and a second matching point pair for each target mask and the standard map mask, wherein the first matching point pair represents the matching point pair between the new positive binary mask of each target mask in the map to be matched and the standard map mask, and the second matching point pair represents the matching point pair between the new inverse binary mask of each target mask in the map to be matched and the standard map mask; and

[0132] The sixth determining module 108 is used to determine a standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each of the target mask maps.

[0133] This application provides a storage medium for storing computer-executable instructions, characterized in that the computer-executable instructions, when executed, implement the steps of the map matching method described in any of the above embodiments.

[0134] It should be noted that the embodiments concerning storage media in this specification and the embodiments concerning blockchain-based service provision methods in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding blockchain-based service provision method described above, and the repeated parts will not be described again.

[0135] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0136] In the 1930s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement to the methodology cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0137] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0138] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0139] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0140] Those skilled in the art will understand that one or more embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0143] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0144] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0145] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0146] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0147] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0148] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0149] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0150] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A map matching method, characterized in that... include: S1. Obtain the map to be matched; S2. Input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps; S3. From the plurality of image segmentation masks, determine a plurality of target masks whose area is greater than a threshold. S4. Perform positive and negative binarization processing on each target mask to obtain a positive binarized mask and a negative binarized mask; S5. For each standard map in the preset standard map library, obtain the standard map mask and the size of the standard map mask; S6. For each standard map in the preset standard map library, generate a new positive binary mask and a new negative binary mask for each target mask based on the standard map mask and the dimensions; wherein, the dimensions represent the number of pixels in height and width, and adjust the pixel dimensions of the positive binary mask and the negative binary mask to be consistent with the pixel dimensions of the standard map mask while maintaining their aspect ratio; S7. Based on an image matching model, match the generated new positive binary mask and the new negative binary mask with the standard map mask respectively, and output them respectively. S8. First matching point pairs and second matching point pairs are generated for each of the target mask images and the standard map mask image, wherein the first matching point pair represents the matching point pair between the new positive binarized mask image of each of the target mask images in the map to be matched and the standard map mask image, and the second matching point pair represents the matching point pair between the new inverse binarized mask image of each of the target mask images in the map to be matched and the standard map mask image; and S8. Based on the first matching point pairs and the second matching point pairs of each of the target mask images, a standard map matching the map to be matched is determined.

2. The method according to claim 1, characterized in that, The step of inputting the map to be matched into an image segmentation model to obtain multiple image segmentation masks includes: dividing the map to be matched into a grid to generate multiple cue points; based on the image segmentation model, using the multiple cue points, extracting the image texture and tile features of the map to be matched; and segmenting the map to be matched into multiple image segmentation masks according to the image texture and the tile features.

3. The method according to claim 1, characterized in that, The step of determining multiple target masks with a value greater than a threshold based on the multiple image segmentation masks includes: determining the area of ​​each of the image segmentation masks; sorting the image segmentation masks according to the area; and selecting the image segmentation mask with a value greater than the threshold as the target mask.

4. The method according to claim 1, characterized in that, The step of generating new positive binary masks and new negative binary masks for each target mask image based on the standard map mask image and the size for each standard map in the preset standard map library includes: if the width pixel count and height pixel count of the standard map mask image are different, obtaining the height pixel count and width pixel count of the standard map mask image based on the standard map mask image; determining the minimum pixel count between the height pixel count and the width pixel count in the standard map mask image; obtaining the pixel count of each target mask image respectively; wherein, the pixel count of the target mask image includes the height pixel count and the width pixel count of the target mask image; determining a first proportional adjustment benchmark based on the minimum pixel count in the standard map mask image, the first proportional adjustment benchmark being the height or width of the standard map mask image; determining a first adjustment ratio based on the first proportional adjustment benchmark and the pixel count of each target mask image; and adjusting each target mask image according to the first adjustment ratio. The width and height pixel counts of the positive binary mask and the inverse binary mask of the membrane image are proportionally enlarged or reduced to generate new positive binary mask and new inverse binary mask. If the width and height pixel counts of the standard map mask are the same, the height pixel count and width pixel count of the target mask are obtained according to the target mask. The minimum pixel count between the height pixel count and the width pixel count in the target mask is determined. A second proportional adjustment benchmark is determined based on the minimum pixel count in the target mask, where the second proportional adjustment benchmark is the height or width in the target mask. A second adjustment ratio is determined based on the second proportional adjustment benchmark and the pixel count of each standard map mask. Based on the second adjustment ratio, the width and height pixel counts of the positive binary mask and the inverse binary mask of each target mask are proportionally enlarged or reduced to generate new positive binary mask and new inverse binary mask.

5. The method according to claim 1, characterized in that, The step of matching the generated new positive binary mask and the new inverse binary mask with the standard map mask based on the image matching model and outputting the first and second matching point pairs of the target mask and the standard map mask respectively includes: the image matching model includes a feature detector and a graph neural network model, and the feature detector is used to extract point and line segment features of the new positive binary mask, the new inverse binary mask and the standard map mask; and based on the graph neural network model, the point and line segment features of the new positive binary mask and the new inverse binary mask are matched with the point and line segment features of the standard map mask respectively, and the first and second matching point pairs of the new positive binary mask and the new inverse binary mask with the standard map mask are output respectively.

6. The method according to claim 1, characterized in that, The preset standard map library includes multiple standard maps; each standard map is binarized to obtain multiple standard map masks; and the step of matching the generated new positive binarized mask and the new inverse binarized mask with the standard map masks based on the image matching model and outputting the first matching point pair and the second matching point pair of each target mask and the standard map mask includes: for each standard map mask, determining the size of the current standard map mask; updating the size of the positive binarized mask and the inverse binarized mask according to the size of the current standard map mask to obtain the new positive binarized mask and the new inverse binarized mask; determining the first matching point pair according to the current standard map mask and the new positive binarized mask; and determining the second matching point pair according to the current standard map mask and the new inverse binarized mask.

7. The method according to claim 6, characterized in that, The step of determining the standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each of the target mask maps includes: sorting the first matching point pair and the second matching point pair of each of the standard map mask maps and each of the target mask maps respectively, and outputting the standard map that matches the target mask map the most times.

8. The method according to claim 1, wherein steps S5 and S6 to S8 are performed using a distributed computing framework.

9. A map matching device, characterized in that... include: The first acquisition module is used to acquire the map to be matched; The first determining module is used to input the map to be matched into the image segmentation model to obtain multiple image segmentation mask maps; The second determining module is used to determine, from the plurality of image segmentation masks, a plurality of target masks whose area is greater than a threshold. The third determining module is used to perform positive and negative binarization processing on each of the target mask images to obtain a positive binarized mask image and a negative binarized mask image; The second acquisition module is used to acquire a standard map mask and the size of the standard map mask for each standard map in the preset standard map library; The fourth determining module is used to generate new positive binary masks and new inverse binary masks for each of the target masks based on the standard map mask and the dimensions; wherein the dimensions represent the number of pixels in height and the number of pixels in width, and the pixel dimensions of the positive binary masks and inverse binary masks are adjusted to be consistent with the pixel dimensions of the standard map mask while maintaining their aspect ratio; and the fifth determining module is used to match the generated new positive binary masks and new inverse binary masks with the standard map mask respectively based on an image matching model, and output each of the target masks respectively. The first matching point pair and the second matching point pair of the mask image and the standard map mask image, wherein the first matching point pair represents the matching point pair between the new positive binarized mask image of each target mask image in the map to be matched and the standard map mask image, and the second matching point pair represents the matching point pair between the new inverse binarized mask image of each target mask image in the map to be matched and the standard map mask image; and a sixth determining module, used to determine the standard map that matches the map to be matched based on the first matching point pair and the second matching point pair of each target mask image.

10. A storage medium for storing computer-executable instructions, characterized in that, When executed, the computer-executable instructions implement the steps of the map matching method according to any one of claims 1-7.

Images