阅读说明：本技术 基于激光点云进行车道线识别的方法和系统 (Method and system for recognizing lane lines based on laser point cloud ) 是由 何弢 廖文龙 尤鑫 于 2019-08-23 设计创作，主要内容包括：本发明提供一种基于激光点云进行车道线识别的方法和系统,根据方位角生成能够唯一识别每个激光点云点的点云时间戳；根据位姿时间戳、激光与世界坐标系的相对位置关系,变换到世界坐标系下；遍历世界坐标系下的激光点云,根据行车轨迹和航向对激光点云进行筛选后,得到感兴趣区域点云；根据感兴趣区域点云的点坐标、激光点云的反射强度,生成栅格地图。本发明将点云进行分组,每一组点云都会具有一个对应的位姿,而不是笼统的每一帧点云给定一个位姿,提高了在较高速行车时的精度。根据实际需求过滤筛选掉很大一部分点云数据,并且经过栅格化,车道线信息明显,降低了存储空间和成本,同时大大降低了非感兴趣区域中进行车道误识别的几率。(The invention provides a method and a system for recognizing lane lines based on laser point clouds, wherein a point cloud timestamp capable of uniquely recognizing each laser point cloud point is generated according to an azimuth angle; transforming to a world coordinate system according to the relative position relation of the pose timestamp and the laser and the world coordinate system; traversing laser point clouds under a world coordinate system, and screening the laser point clouds according to the driving track and the course to obtain point clouds of an area of interest; and generating a grid map according to the point coordinates of the point cloud of the region of interest and the reflection intensity of the laser point cloud. The invention groups the point clouds, each group of point clouds has a corresponding pose instead of giving a pose to each frame of point clouds in a general way, thereby improving the precision in high-speed driving. A large part of point cloud data is filtered and screened according to actual requirements, and through rasterization, lane line information is obvious, storage space and cost are reduced, and the probability of lane error recognition in a non-interest area is greatly reduced.)

1. A method for recognizing lane lines based on laser point cloud is characterized by comprising the following steps:

and a timestamp corresponding step: collecting laser point clouds, calculating an azimuth angle of each laser point cloud point, and generating a point cloud timestamp capable of uniquely identifying each laser point cloud point according to the azimuth angle;

point cloud conversion: identifying and obtaining a corresponding pose timestamp according to the point cloud timestamp, and transforming the laser point cloud to a world coordinate system according to the pose timestamp and the relative position relation between the laser and the world coordinate system;

point cloud screening: traversing laser point clouds under a world coordinate system, and screening the laser point clouds according to the driving track and the course to obtain point clouds of an area of interest;

point cloud grid step: and generating a grid map according to the point coordinates of the point cloud of the region of interest and the reflection intensity of the laser point cloud.

2. The method of claim 1, wherein the timestamp mapping step comprises:

the collection step comprises: acquiring original point cloud data through scanning of a laser radar, recording points with point cloud values being infinite values and distances between the points and the laser radar exceeding a set distance in the original point cloud data as invalid points, and removing the invalid points to obtain laser point cloud;

grouping: calculating azimuth angles of all points in the laser point cloud, sequencing the azimuth angles, taking the difference value of the maximum azimuth angle and the minimum azimuth angle as a frame scanning range, grouping according to the frame scanning range, setting an initial timestamp, adding a time interval corresponding to a grouped group number to the initial timestamp, and recording as the grouped point cloud timestamp.

3. The method for lane line identification based on laser point cloud of claim 1, wherein said point cloud conversion step comprises:

pose matching: traversing the point cloud timestamps, searching two corresponding pose timestamps according to the point cloud timestamps of each group, and recording the two bit sub-timestamps as the pose timestamps corresponding to the point cloud timestamps after interpolation;

and (3) coordinate transformation: and transforming the laser point cloud to the world coordinate system according to the set relative position relation between the laser radar and the world coordinate system and by combining the pose timestamp.

4. The method of claim 1, wherein the point cloud screening step comprises:

and (3) position screening: removing laser point clouds of which the positive distance of a Y axis is smaller than a first set width and the negative distance of the Y axis is smaller than a second set width from the laser point clouds under the world coordinate system;

a height screening step: and removing the laser point cloud with the height greater than the set height from the ground for the laser point cloud under the world coordinate system.

5. The method for lane line identification based on laser point cloud of claim 1, wherein said point cloud gridding step comprises:

and (3) determining the size: traversing the point cloud of the region of interest, and prefabricating the length and width dimensions of the grid map;

determining the resolution: determining the resolution of the grid map according to the length, the width and the grid resolution;

determining the position: determining the position of a point in the point cloud of the region of interest in a grid map, recording the intensity value of the point as the corresponding value of the point, and taking the maximum value of the intensity values in all the points contained in one grid as the grid point cloud value of the grid;

generating a map: and generating a grid map according to the grid point cloud value of each grid.

6. A system for lane line identification based on laser point cloud, comprising:

a timestamp correspondence module: collecting laser point clouds, calculating an azimuth angle of each laser point cloud point, and generating a point cloud timestamp capable of uniquely identifying each laser point cloud point according to the azimuth angle;

a point cloud conversion module: identifying and obtaining a corresponding pose timestamp according to the point cloud timestamp, and transforming the laser point cloud to a world coordinate system according to the pose timestamp and the relative position relation between the laser and the world coordinate system;

a point cloud screening module: traversing laser point clouds under a world coordinate system, and screening the laser point clouds according to the driving track and the course to obtain point clouds of an area of interest;

a point cloud grid module: and generating a grid map according to the point coordinates of the point cloud of the region of interest and the reflection intensity of the laser point cloud.

7. The system for lane line identification based on a laser point cloud of claim 6, wherein said timestamp correspondence module comprises:

an acquisition module: acquiring original point cloud data through scanning of a laser radar, recording points with point cloud values being infinite values and distances between the points and the laser radar exceeding a set distance in the original point cloud data as invalid points, and removing the invalid points to obtain laser point cloud;

a grouping module: calculating azimuth angles of all points in the laser point cloud, sequencing the azimuth angles, taking the difference value of the maximum azimuth angle and the minimum azimuth angle as a frame scanning range, grouping according to the frame scanning range, setting an initial timestamp, adding a time interval corresponding to a grouped group number to the initial timestamp, and recording as the grouped point cloud timestamp.

8. The system for lane line identification based on a laser point cloud of claim 6, wherein said point cloud conversion module comprises:

a pose matching module: traversing the point cloud timestamps, searching two corresponding pose timestamps according to the point cloud timestamps of each group, and recording the two bit sub-timestamps as the pose timestamps corresponding to the point cloud timestamps after interpolation;

a coordinate transformation module: and transforming the laser point cloud to the world coordinate system according to the set relative position relation between the laser radar and the world coordinate system and by combining the pose timestamp.

9. The system for lane line identification based on laser point cloud of claim 6, wherein said point cloud screening module comprises:

a position screening module: removing laser point clouds of which the positive distance of a Y axis is smaller than a first set width and the negative distance of the Y axis is smaller than a second set width from the laser point clouds under the world coordinate system;

a height screening module: and removing the laser point cloud with the height greater than the set height from the ground for the laser point cloud under the world coordinate system.

10. The system for lane line identification based on a laser point cloud of claim 6, wherein said point cloud grid module comprises:

a sizing module: traversing the point cloud of the region of interest, and prefabricating the length and width dimensions of the grid map;

a resolution determination module: determining the resolution of the grid map according to the length, the width and the grid resolution;

a determine location module: determining the position of a point in the point cloud of the region of interest in a grid map, recording the intensity value of the point as the corresponding value of the point, and taking the maximum value of the intensity values in all the points contained in one grid as the grid point cloud value of the grid;

a map generation module: and generating a grid map according to the grid point cloud value of each grid.