A vision-assisted lidar localisation system is proposed to address the problems such as long reposi-tioning time or even failure of repositioning for single-wire lidar.Firstly,a visual sensor with richer infor-mation is introduced to link feature maps and raster maps,and visual repositioning is preferentially used to find the rough position and pass it into the laser positioning system,which will do the fine repositioning;then,for the corridor-like environment where the single-line laser positioning system is prone to causing the positioning drift,a windowed straight-line fitting method is proposed for identifying the corridor-like environment so as to improve the odometry.Finally,for the possible incompleteness of single-line radar point cloud,the best-fit sub-map is used to replace the nearest sub-map for the laser fine repositioning task.Experiments show that the system can effectively extract environmental line features to identify corridor-like environments,and can effectively reduce the matching time and improve the correct matching rate to i-dentify the correct position of the robot when fighting against similar environments.
关键词
移动机器人/2D激光雷达/直线拟合/视觉定位/重定位
Key words
mobile robot/2D lidar/linear fitting/visual localisation/repositioning
维普
万方数据