Road Semantic Point Cloud Map Vectorization Method, System and Storage Medium

Abstract

A road semantic point cloud map vectorization method, system, and storage medium segment the preprocessed data to obtain M point cloud subgraphs and M local trajectories, then establish M spatial auxiliary coordinates, and transform the point cloud subsets into spatial auxiliary coordinates for segmented vectorization to obtain M two-dimensional vector subgraphs. These subgraphs are then spliced and fused in a unified coordinate system to output a complete vector map, achieving rapid two-dimensional vectorization of segments and coordinate systems, reducing the vectorization cost of road semantic point cloud maps and improving their vectorization efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of high-precision map technology, and specifically relates to a road semantic point cloud map vectorization method, system and storage medium. Background Art

[0002] Semantic point cloud maps are point cloud data with semantic information. They are a type of three-dimensional scene information derived from semantic segmentation and point cloud fusion. They can be obtained through technologies such as visual SLAM (Simultaneous Localization and Mapping) and LiDAR. Point cloud map data is large and inconvenient for spatial retrieval and analysis. Vector map data, on the other hand, has a complex and compact structure with low redundancy, making it easier to transmit, store, and perform spatial retrieval and analysis.

[0003] Chinese patent publication number CN113378800A discloses a method for automatic road marking classification and vectorization. This method uses collected vehicle trajectory information to segment a vehicle-mounted laser point cloud into multiple point clouds according to defined lengths and widths. Each point cloud is then classified and ground points extracted. Road markings are then extracted from the ground points. Finally, road markings are classified and vectorized according to a template. This method solves the problem of vectorizing local ground markings, but it is not suitable for vectorizing large-scale, full-element road point clouds. Furthermore, the trajectory-based segmentation method is not applicable to overlapping areas, such as overpasses.

[0004] Therefore, it is necessary to develop a road semantic point cloud map vectorization method to meet the needs of road point cloud vectorization processing of large scenes and full elements when performing road semantic point cloud vectorization, and to solve problems such as the low degree of automation in the vectorization process and incomplete elements of the vectorization results. Summary of the Invention

[0005] The road semantic point cloud map vectorization method disclosed in the present invention can automatically vectorize the road semantic point cloud map of all elements, reduce the road semantic point cloud map vectorization cost, and improve the road semantic point cloud map vectorization efficiency.

[0006] The present invention also discloses a system for vectorizing road semantic point cloud maps, which can automatically vectorize road semantic point cloud maps of all elements, reduce the cost of road semantic point cloud map vectorization, and improve the efficiency of road semantic point cloud map vectorization.

[0007] The present invention also discloses a storage medium that can automatically vectorize a road semantic point cloud map with all elements, reduce the vectorization cost of the road semantic point cloud map, and improve the vectorization efficiency of the road semantic point cloud map.

[0008] The road semantic point cloud map vectorization method disclosed in the present invention includes the following steps:

[0009] Step 1: Obtain target area data;

[0010] Step 2: preprocessing the target area data to obtain preprocessed data;

[0011] Step 3: Segment the preprocessed data to obtain M point cloud subgraphs and M local trajectories; where M is a positive integer;

[0012] Step 4: determining M spatial auxiliary coordinate systems according to the M local trajectories, and vectorizing the point sets in the M point cloud subgraphs to obtain M vector subgraphs;

[0013] Step 5: converting the M vector subgraphs into vector subgraphs in a local unified coordinate system;

[0014] Step 6: Splice and fuse the M vector sub-maps in the local unified coordinate system to output a complete vector map.

[0015] Furthermore, in step 4, M spatial auxiliary coordinate systems are determined according to the M local trajectories, and conversion parameters from the local unified coordinate system to the spatial auxiliary coordinate systems are calculated;

[0016] The three-dimensional coordinates of the point cloud data corresponding to the point cloud subgraphs in the M spatial auxiliary coordinate systems are calculated respectively, and then projected onto the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets; the M two-dimensional point sets are vectorized respectively to obtain M two-dimensional vector subgraphs.

[0017] Furthermore, in step 4, the calculation of M two-dimensional point sets includes the following sub-steps:

[0018] Step S401: performing straight line fitting on the M segments of local trajectories to obtain M auxiliary straight lines;

[0019] Step S402: Obtain the rotation matrix R and the translation vector t according to the origin and coordinate axis definitions of the M spatial auxiliary coordinate systems;

[0020] Step S403: point cloud data projection. Calculate the three-dimensional coordinates of the corresponding point cloud data in the M spatial auxiliary coordinate systems respectively, and project the three-dimensional coordinates onto the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets. The projection calculation formula is as follows:

[0021]

[0022] Wherein, (x′, y′, z′) is the coordinate in the auxiliary spatial coordinate system, (x, y, z) is the coordinate in the local unified coordinate system, and the two-dimensional coordinate (x′, y′) is obtained after being projected onto the XOY plane.

[0023] Furthermore, in step 5, the M two-dimensional vector subgraphs are converted into M three-dimensional vector subgraphs;

[0024] In step 6, the M three-dimensional vector sub-images are spliced and fused to output a complete vector map.

[0025] Furthermore, the data segmentation in step 3 includes the following sub-steps:

[0026] Step S301 constructs empty trajectory point sequences V1 and V2;

[0027] Step S302: inserting all the trajectory points in the trajectory data into the trajectory point sequence V1;

[0028] Step S303: extract the trajectory point with the smallest timestamp from the trajectory point sequence V1 and insert it into the trajectory point sequence V2;

[0029] Step S304 calculates the trajectory length l of the trajectory point sequence V2. The trajectory length calculation formula is as follows:

[0030]

[0031] Where, P i represents the i-th trajectory point in the trajectory point sequence V2, n represents the number of trajectory points in the trajectory point sequence V2, |P i -P i-1 | represents the distance between the i-1th trajectory point and the i-th trajectory point;

[0032] Step S305 compares the trajectory length l with a preset length threshold L. If l<L, go to step S303;

[0033] Step S306: Divide all the trajectory points in the trajectory point sequence V2 into a local trajectory;

[0034] Step S307 divides the point cloud data corresponding to the start and end time range of the trajectory point sequence V2 into a point cloud subgraph, and clears the trajectory point sequence V2;

[0035] Step S308 determines whether all the track points in V1 have been retrieved; if not, return to step S303; if all have been retrieved, end.

[0036] Furthermore,

[0037] In step 4, the origin of the spatial auxiliary coordinate system is the foot point of the trajectory point with the smallest timestamp in the local trajectory on the auxiliary straight line; the direction of the X-axis of the spatial auxiliary coordinate system points from the origin of the spatial auxiliary coordinate system to the foot point of the trajectory point with the largest timestamp in the local trajectory on the auxiliary straight line; the Y-axis of the spatial auxiliary coordinate system is parallel to the XOY plane of the local unified coordinate system and perpendicular to the X-axis of the spatial auxiliary coordinate system; the Z-axis of the spatial auxiliary coordinate system passes through the origin of the spatial auxiliary coordinate system and is perpendicular to the X-axis and Y-axis of the spatial auxiliary coordinate system at the same time.

[0038] Furthermore, in step 2, the target area data is preprocessed, specifically including the following sub-steps:

[0039] Step S201: unifying the spatiotemporal reference base, converting the point cloud data in the semantic point cloud map and the synchronized trajectory data to the same local unified coordinate system base;

[0040] Step S202: removing outliers from the point cloud data, performing statistical filtering on the point cloud data in the semantic point cloud map for each semantic category;

[0041] In step S203 , the point cloud data is downsampled, and voxel filtering is performed on the point cloud data in the semantic point cloud map for each semantic category.

[0042] Furthermore, in step 5, the two-dimensional vector sub-graph is converted into a three-dimensional vector sub-graph according to the conversion parameters. The calculation formula is as follows:

[0043]

[0044] Among them, (x′, y′) is the geometric coordinate of the two-dimensional vector object in the XOY plane of the spatial auxiliary coordinate system, and (x, y, z) is the three-dimensional coordinate of the vector object in the local unified coordinate system.

[0045] The present invention also discloses a system for vectorizing a road semantic point cloud map, comprising a memory and a controller, wherein the memory stores a computer-readable program, and the computer-readable program can execute a method for vectorizing a road semantic point cloud map when called by the controller.

[0046] The present invention also discloses a storage medium storing a computer-readable program, which can execute a road semantic point cloud map vectorization method when called.

[0047] The beneficial technical effects of the present invention are as follows: a road semantic point cloud map vectorization method, system and storage medium segments the preprocessed data to obtain M point cloud subgraphs and M local trajectories, then establishes M spatial auxiliary coordinates, and converts the point cloud subsets into spatial auxiliary coordinates for segmented vectorization to obtain M two-dimensional vector subgraphs, which are then spliced and fused under a unified coordinate system to output a complete vector map, thereby realizing fast two-dimensional vectorization of segmented and sub-coordinate systems, reducing the vectorization cost of road semantic point cloud maps, and improving the vectorization efficiency of road semantic point cloud maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the road semantic point cloud map vectorization method of the present invention.

[0049] Figure 2 This is a data preprocessing flowchart for the road semantic point cloud map vectorization method of the present invention.

[0050] Figure 3 This is a flowchart of data segmentation of the road semantic point cloud map vectorization method of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be described in detail below with reference to the accompanying drawings. Figures 1 to 3 As shown, the road semantic point cloud map vectorization method includes the following steps:

[0052] Step 1: Obtain target area data;

[0053] Step 2 preprocesses the target area data to obtain preprocessed data; the target area data is preprocessed, specifically including the following sub-steps:

[0054] Step S201: unifying the spatiotemporal reference base, converting the point cloud data in the semantic point cloud map and the synchronized trajectory data to the same local unified coordinate system base;

[0055] Step S202: removing outliers from the point cloud data, performing statistical filtering on the point cloud data in the semantic point cloud map for each semantic category;

[0056] In step S203 , the point cloud data is downsampled, and voxel filtering is performed on the point cloud data in the semantic point cloud map for each semantic category.

[0057] Step 3 segments the pre-processed data to obtain M point cloud sub-graphs and M local trajectories; wherein M is a positive integer; the data segmentation in step 3 includes the following sub-steps:

[0058] Step S301 constructs empty trajectory point sequences V1 and V2;

[0059] Step S302: inserting all the trajectory points in the trajectory data into the trajectory point sequence V1;

[0060] Step S303: extract the trajectory point with the smallest timestamp from the trajectory point sequence V1 and insert it into the trajectory point sequence V2;

[0061] Step S304 calculates the trajectory length l of the trajectory point sequence V2. The trajectory length calculation formula is as follows:

[0062]

[0063] Where, P i represents the i-th trajectory point in the trajectory point sequence V2, n represents the number of trajectory points in the trajectory point sequence V2, |P i -P i-1 | represents the distance between the i-1th trajectory point and the i-th trajectory point;

[0064] Step S305 compares the trajectory length l with a preset length threshold L. If l<L, go to step S303;

[0065] Step S306: Divide all the trajectory points in the trajectory point sequence V2 into a local trajectory;

[0066] Step S307 divides the point cloud data corresponding to the start and end time range of the trajectory point sequence V2 into a point cloud subgraph, and clears the trajectory point sequence V2;

[0067] Step S308 determines whether all the track points in V1 have been retrieved; if not, return to step S303; if all have been retrieved, end.

[0068] Step 4 determines M spatial auxiliary coordinate systems based on the M local trajectories, and vectorizes the point sets in the M point cloud subgraphs to obtain M vector subgraphs. Specifically, M spatial auxiliary coordinate systems are determined based on the M local trajectories, and conversion parameters from the local unified coordinate system to the spatial auxiliary coordinate systems are calculated; the three-dimensional coordinates of the point cloud data corresponding to the point cloud subgraphs in the M spatial auxiliary coordinate systems are calculated respectively, and then projected onto the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets; and the M two-dimensional point sets are vectorized to obtain M two-dimensional vector subgraphs.

[0069] Determination of spatial auxiliary coordinate system:

[0070] The origin of the spatial auxiliary coordinate system is the foot point of the trajectory point with the smallest timestamp in the local trajectory on the auxiliary straight line; the direction of the X-axis of the spatial auxiliary coordinate system points from the origin of the spatial auxiliary coordinate system to the foot point of the trajectory point with the largest timestamp in the local trajectory on the auxiliary straight line; the Y-axis of the spatial auxiliary coordinate system is parallel to the XOY plane of the local unified coordinate system and perpendicular to the X-axis of the spatial auxiliary coordinate system; the Z-axis of the spatial auxiliary coordinate system passes through the origin of the spatial auxiliary coordinate system and is perpendicular to the X-axis and Y-axis of the spatial auxiliary coordinate system at the same time.

[0071] The calculation of M two-dimensional point sets includes the following sub-steps:

[0072] Step S401 performs linear fitting on the M segments of local trajectories to obtain M auxiliary straight lines. The linear fitting method is the least squares method, and the auxiliary straight lines are expressed in a point-to-point form. The linear expression is as follows:

[0073]

[0074] Where x0, y0, z0, a, b, and c are the parameters of the straight line equation.

[0075] Step S402 obtains the rotation matrix R and the translation vector t according to the origin and coordinate axis definitions of the M spatial auxiliary coordinate systems.

[0076] Step S403: point cloud data projection. Calculate the three-dimensional coordinates of the corresponding point cloud data in the M spatial auxiliary coordinate systems respectively, and project the three-dimensional coordinates onto the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets. The projection calculation formula is as follows:

[0077]

[0078] Wherein, (x′, y′, z′) is the coordinate in the auxiliary spatial coordinate system, (x, y, z) is the coordinate in the local unified coordinate system, and the two-dimensional coordinate (x′, y′) is obtained after being projected onto the XOY plane.

[0079] The steps of vectorizing the M two-dimensional point sets are as follows: performing vector extraction on the M two-dimensional point sets to obtain M two-dimensional vector subgraphs. The vector extraction includes point vector extraction, line vector extraction, and surface vector extraction.

[0080] Step 5: converting the M vector subgraphs into vector subgraphs in a local unified coordinate system;

[0081] Specifically, M two-dimensional vector subgraphs are converted into M three-dimensional vector subgraphs. According to the conversion parameters, the two-dimensional vector subgraphs are converted into three-dimensional vector subgraphs. The calculation formula is as follows:

[0082]

[0083] Among them, (x′, y′) is the geometric coordinate of the two-dimensional vector object in the XOY plane of the spatial auxiliary coordinate system, and (x, y, z) is the three-dimensional coordinate of the vector object in the local unified coordinate system.

[0084] Step 6: Splice and fuse the M vector sub-images in the local unified coordinate system to output a complete vector map. Specifically, splice and fuse the M three-dimensional vector sub-images to output a complete vector map. Splicing and fusing to output a complete vector map includes the following sub-steps:

[0085] The line and surface vector objects between adjacent sub-images in the M three-dimensional vector sub-images are spliced; and the point vector objects between adjacent sub-images in the M three-dimensional vector sub-images are fused.

[0086] The present invention also discloses a system for vectorizing a road semantic point cloud map, comprising a memory and a controller, wherein the memory stores a computer-readable program, and the computer-readable program can execute a method for vectorizing a road semantic point cloud map when called by the controller.

[0087] The present invention also discloses a storage medium storing a computer-readable program, which can execute a road semantic point cloud map vectorization method when called.

Claims

1. A road semantic point cloud map vectorization method, characterized by: The following steps are included: Step 1: Obtain target area data; Step 2: preprocessing the target area data to obtain preprocessed data; Step 3: Segment the preprocessed data to obtain M semantic point cloud subgraphs and M local trajectories; where M is a positive integer; Step 4: Determine M spatial auxiliary coordinate systems based on the M local trajectories, and calculate the conversion parameters from the local unified coordinate system to the spatial auxiliary coordinate system; calculate the three-dimensional coordinates of the point cloud data of the corresponding semantic point cloud subgraphs in the M spatial auxiliary coordinate systems respectively, and then project them to the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets; vectorize the M two-dimensional point sets respectively to obtain M two-dimensional vector subgraphs; the origin of the spatial auxiliary coordinate system is the foot point of the trajectory point with the smallest timestamp in the local trajectory on the auxiliary straight line, and the auxiliary straight line is obtained by linear fitting of the M local trajectories respectively; the direction of the X-axis of the spatial auxiliary coordinate system is from the origin of the spatial auxiliary coordinate system to the foot point of the trajectory point with the largest timestamp in the local trajectory on the auxiliary straight line; the Y-axis of the spatial auxiliary coordinate system is parallel to the XOY plane of the local unified coordinate system and perpendicular to the X-axis of the spatial auxiliary coordinate system; the Z-axis of the spatial auxiliary coordinate system passes through the origin of the spatial auxiliary coordinate system and is perpendicular to both the X-axis and the Y-axis of the spatial auxiliary coordinate system; Step 5: converting the M vector subgraphs into vector subgraphs in a local unified coordinate system; Step 6: Splice and fuse the M vector sub-maps in the local unified coordinate system to output a complete vector map.

2. The road semantic point cloud map vectorization method according to claim 1, characterized in that: In step 4, the calculation of M two-dimensional point sets includes the following sub-steps: Step S401: performing straight line fitting on the M segments of local trajectories to obtain M auxiliary straight lines; Step S402: Obtain the rotation matrix R and the translation vector t according to the origin and coordinate axis definitions of the M spatial auxiliary coordinate systems; Step S403: point cloud data projection. Calculate the three-dimensional coordinates of the corresponding point cloud data in the M spatial auxiliary coordinate systems respectively, and project the three-dimensional coordinates onto the XOY plane of the spatial auxiliary coordinate system to obtain M two-dimensional point sets. The projection calculation formula is as follows: Wherein, (x′, y′, z′) is the coordinate in the auxiliary spatial coordinate system, (x, y, z) is the coordinate in the local unified coordinate system, and the two-dimensional coordinate (x′, y′) is obtained after being projected onto the XOY plane.

3. The road semantic point cloud map vectorization method according to claim 1, characterized in that: In step 5, the M two-dimensional vector subgraphs are converted into M three-dimensional vector subgraphs; In step 6, the M three-dimensional vector sub-images are spliced and fused to output a complete vector map.

4. The method for vectorizing a road semantic point cloud map according to any one of claims 1 to 3, wherein: Data segmentation in step 3 includes the following sub-steps: Step S301 constructs empty trajectory point sequences V1 and V2; Step S302 inserts all trajectory points in the trajectory data into the trajectory point sequence V1; Step S303: extract the trajectory point with the smallest timestamp from the trajectory point sequence V1 and insert it into the trajectory point sequence V2; Step S304 calculates the trajectory length l of the trajectory point sequence V2. The trajectory length calculation formula is as follows: Where, P i represents the i-th trajectory point in the trajectory point sequence V2, n represents the number of trajectory points in the trajectory point sequence V2, |P i -P i-1 | represents the distance between the i-1th trajectory point and the i-th trajectory point; Step S305 compares the trajectory length l with a preset length threshold L. If l<L, go to step S303; Step S306: Divide all the trajectory points in the trajectory point sequence V2 into a local trajectory; Step S307 divides the point cloud data corresponding to the start and end time range of the trajectory point sequence V2 into a point cloud subgraph, and clears the trajectory point sequence V2; Step S308 determines whether all the track points in V1 have been retrieved; if not, return to step S303; if all have been retrieved, end.

5. The method for vectorizing a road semantic point cloud map according to any one of claims 1 to 3, wherein: In step 2, the target area data is preprocessed, which specifically includes the following sub-steps: Step S201: unifying the spatiotemporal reference base, converting the point cloud data and the synchronized trajectory data in the semantic point cloud map to the same local unified coordinate system base; Step S202: removing outliers from the point cloud data, performing statistical filtering on the point cloud data in the semantic point cloud map for each semantic category; In step S203 , the point cloud data is downsampled, and voxel filtering is performed on the point cloud data in the semantic point cloud map for each semantic category.

6. The road semantic point cloud map vectorization method according to claim 3, characterized in that: In step 5, the two-dimensional vector sub-graph is converted into a three-dimensional vector sub-graph according to the conversion parameters. The calculation formula is as follows: Among them, (x′, y′) is the geometric coordinate of the two-dimensional vector object in the XOY plane of the spatial auxiliary coordinate system, and (x, y, z) is the three-dimensional coordinate of the vector object in the local unified coordinate system.

7. A system for vectorizing road semantic point cloud maps, characterized by: The method comprises a memory and a controller, wherein the memory stores a computer-readable program, and when the computer-readable program is called by the controller, the method can execute the road semantic point cloud map vectorization method as described in any one of claims 1 to 6.

8. A storage medium, characterized in that: A computer-readable program is stored therein, and when the computer-readable program is called, it can execute the road semantic point cloud map vectorization method as described in any one of claims 1 to 6.

Images