CN-122024192-A - Method and device for detecting road boundary and unmanned vehicle

CN122024192ACN 122024192 ACN122024192 ACN 122024192ACN-122024192-A

Abstract

The application discloses a method and a device for detecting a road boundary and an unmanned vehicle. The method comprises the steps of obtaining point cloud data of a working area and original track data of a vehicle, wherein the original track data are track data obtained when the vehicle runs on a road in the working area, respectively projecting the point cloud data and the original track data from a three-dimensional space to a two-dimensional space to obtain a projection image and target track data, extracting a first road boundary of the road from the projection image based on the target track data, and mapping the first road boundary from the two-dimensional space to the three-dimensional space to obtain the target road boundary of the road. The application solves the technical problem of lower accuracy of road boundary detection in the related technology.

Inventors

SHI YUNFEI
JIANG CHENG

Assignees

易控智驾科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260127

Claims (10)

1. A method of detecting a road boundary, comprising: Acquiring point cloud data of a working area and original track data of a vehicle, wherein the original track data are track data obtained by the vehicle driving on a road in the working area; Respectively projecting the point cloud data and the original track data from a three-dimensional space to a two-dimensional space to obtain a projection image and target track data; Extracting a first road boundary of the road from the projection image based on the target track data; And mapping the first road boundary from a two-dimensional space to a three-dimensional space to obtain a target road boundary of the road.
2. The method of claim 1, wherein extracting a first road boundary of the road from the projection image based on the target trajectory data comprises: Extracting a sub-projection image from the projection image based on the attributes corresponding to different points in the point cloud data, wherein the attributes of the points corresponding to the sub-projection image are preset attributes; Extracting the outline of the sub-projection image to obtain an initial road boundary of the road; And screening the initial road boundary based on the target track data to obtain the first road boundary.
3. The method of claim 2, wherein the screening the initial road boundary based on the target trajectory data to obtain the first road side boundary comprises: Generating a pixel template aiming at any track point in the target track data based on the two-dimensional position information of the track point, wherein the pixel template is a pixel set of a preset shape edge taking the track point as the center; Constructing a plurality of ray segments by taking the track point as a starting point and a plurality of edge points on the pixel template as end points; Determining the intersection points of the plurality of ray segments and the initial road boundary to obtain candidate boundary points corresponding to the track points; And obtaining the first road boundary based on candidate boundary points corresponding to the plurality of track points in the target track data.
4. The method of claim 2 or 3, wherein the projected image comprises a plurality of images, the different images having different projection resolutions, and wherein the screening the initial road boundary based on the target trajectory data to obtain the first road boundary comprises: screening the initial road boundaries corresponding to the images based on the target track data respectively to obtain a plurality of screened road boundaries; Weighting the plurality of screened road boundaries to obtain the first road boundary; preferably, the weighting processing is performed on the plurality of screened road boundaries to obtain the first road boundary, including: querying the plurality of screened road boundaries aiming at any one track point in the target track data to obtain a query result, wherein the query result is used for representing whether candidate boundary points corresponding to the track points exist in the plurality of screened road boundaries; determining weight values corresponding to the plurality of screened road boundaries based on the query result, wherein the sum of the weight values corresponding to at least one road boundary to be weighted in the plurality of screened road boundaries is 1, the weight values corresponding to other road boundaries are 0, candidate boundary points corresponding to the track points exist in the at least one road boundary to be weighted, and the other road boundaries are screened road boundaries except for the at least one road boundary to be weighted in the plurality of screened road boundaries; and weighting the plurality of filtered road boundaries based on the weight values corresponding to the plurality of filtered road boundaries to obtain the first road boundary.
5. The method according to claim 2, wherein the performing contour extraction on the sub-projection image to obtain an initial road boundary of the road comprises: morphological processing is carried out on the sub-projection images to obtain processed images; and extracting the outline of the processed image to obtain the initial road boundary.
6. The method of claim 5, wherein the projection image comprises a plurality of images, the projection resolutions of the different images are different, and the performing contour extraction on the processed image to obtain the initial road boundary comprises: respectively extracting contours of images corresponding to different projection resolutions to obtain a plurality of road boundaries; And merging the plurality of road boundaries to obtain the initial road boundary.
7. The method according to any one of claims 1 to 6, wherein projecting the point cloud data from a three-dimensional space to a two-dimensional space to obtain a projection image, comprises: Determining size information of the projection image based on three-dimensional position information of a plurality of points in the point cloud data and target projection resolution; determining projection position information of any point in the projection image based on the three-dimensional position information of the point, the target projection resolution and the position information of the points in the point cloud data; obtaining the projection image based on the attribute and the projection position information of the plurality of points; Preferably, the method further comprises: Determining index information of the point based on the projection position information of the point; And storing the three-dimensional position information of the point and the index information of the point to a target data structure correspondingly.
8. The method according to any one of claims 1 to 6, wherein mapping the first road boundary from the two-dimensional space to the three-dimensional space to obtain a target road boundary of the road comprises: determining target index information of any boundary point in the first road boundary based on the two-dimensional position information of the boundary point; Determining three-dimensional position information of the boundary point from a target data structure based on the target index information; And obtaining the target road boundary based on the three-dimensional position information of a plurality of boundary points in the first road boundary.
9. An apparatus for detecting a road boundary, comprising: The acquisition module is used for acquiring point cloud data of a working area and original track data of a vehicle, wherein the original track data are track data obtained by the vehicle driving on a road in the working area; The first projection module is used for respectively projecting the point cloud data and the original track data from a three-dimensional space to a two-dimensional space to obtain a projection image and target track data; The extraction module is used for extracting a first road boundary of the road from the projection image based on the target track data; and the second projection module is used for mapping the first road boundary from a two-dimensional space to a three-dimensional space to obtain a target road boundary of the road.
10. An unmanned vehicle, comprising: A memory storing an executable program; A processor for executing the program, wherein the program when run performs the method of any one of claims 1 to 8.

Description

Method and device for detecting road boundary and unmanned vehicle Technical Field The application relates to the technical fields of automatic driving technology and data processing, in particular to a method and a device for detecting road boundaries and an unmanned vehicle. Background In special working areas such as mine roads, the lack of standardized human identification and guidance features constitutes a unique environmental challenge, which directly makes the traditional road detection technology based on significant visual features difficult to effectively apply. In the related art, although the vision solution is dominant, performance of the vision system in road boundary detection is seriously hindered due to unstable mine illumination conditions, shielding of dust, lack of contrast of the background, and the like. In addition, even if a device such as a laser radar is introduced to acquire point cloud data in an attempt to extract a road boundary therefrom, the quality of the obtained data is poor under natural conditions without obvious features, resulting in lower accuracy of road boundary detection in the related art. In view of the above problems, no effective solution has been proposed at present. Disclosure of Invention The embodiment of the application provides a method and a device for detecting a road boundary and an unmanned vehicle, which are used for at least solving the technical problem of lower accuracy of road boundary detection in the related technology. According to one aspect of the embodiment of the application, a method for detecting a road boundary is provided, which comprises the steps of obtaining point cloud data of a working area and original track data of a vehicle, wherein the original track data are track data obtained by the vehicle driving on a road in the working area, respectively projecting the point cloud data and the original track data from a three-dimensional space to a two-dimensional space to obtain a projection image and target track data, extracting a first road boundary of the road from the projection image based on the target track data, and mapping the first road boundary from the two-dimensional space to the three-dimensional space to obtain the target road boundary of the road. Further, extracting a first road boundary of the road from the projection image based on the target track data, wherein the first road boundary comprises extracting a sub-projection image from the projection image based on attributes corresponding to different points in the point cloud data, and the attributes of the points corresponding to the sub-projection image are preset attributes; and screening the initial road boundary based on the target track data to obtain a first road boundary. The method comprises the steps of selecting an initial road boundary based on target track data to obtain a first road boundary, generating a pixel template based on two-dimensional position information of any track point in the target track data, wherein the pixel template is a pixel set of edges of a preset shape taking the track point as a center, constructing a plurality of ray segments by taking the track point as a starting point and a plurality of edge points on the pixel template as end points, determining the intersection points of the plurality of ray segments and the initial road boundary to obtain candidate boundary points corresponding to the track point, and obtaining the first road boundary based on the candidate boundary points corresponding to the plurality of track points in the target track data. Further, the projection image comprises a plurality of images, and the projection resolutions corresponding to different images are different; the method comprises the steps of screening initial road boundaries based on target track data to obtain first road boundaries, screening the initial road boundaries corresponding to a plurality of images based on the target track data to obtain a plurality of screened road boundaries, weighting the plurality of screened road boundaries to obtain the first road boundaries, preferably weighting the plurality of screened road boundaries to obtain the first road boundaries, inquiring the plurality of screened road boundaries according to any track point in the target track data to obtain inquiry results, wherein the inquiry results are used for representing whether candidate boundary points corresponding to the track points exist in the plurality of screened road boundaries, determining weight values corresponding to the plurality of screened road boundaries based on the inquiry results, wherein the sum of the weight values corresponding to at least one road boundary to be weighted is 1, the weight values corresponding to other road boundaries are 0, the candidate boundary points corresponding to the track points exist in at least one road boundary to be weighted, the other road boundaries are the candidate boundary points corresponding to t