1. A laser SLAM method for unmanned vehicles in point cloud degenerated tunnel environments
- Author
-
LI Shuaixin, LI Jiuren, TIAN Bin, CHEN Long, WANG Li, and LI Guangyun
- Subjects
laser slam ,point feature extraction ,point cloud registration ,Mathematical geography. Cartography ,GA1-1776 ,point cloud degeneracy - Abstract
Laser SLAM enables to locate the vehicle itself even in an unknown environment, and to efficiently sample the three-dimensional geospatial information of the traversed environment, which has been drawn wide attention in the field of autonomous driving in recent years. To improve the accuracy and performance of the laser SLAM system in a point cloud degenerated tunnel environment, we present an intensity enhanced laser SLAM approach based on LOAM. First, we improve the feature extraction module of LOAM. An adaptive feature extraction method based on spherical projection image is presented to extract line, façde, ground and reflectors from a single laser sweep. Besides, to solve the issue on point cloud registration degeneracy in tunnel environments, we presented intensity feature-basedregistration approach to fix the vehicle pose resulting from the geometric feature-based registration error. Reflecting features in the surrounding are adaptively extracted to ensure the adaptivity of our improved laser SLAM approach. The experimental results show that the proposed method presented the better and more robust performance especially in degenerated environments, e.g., long straight tunnel, comparing with the performance of LOAM and HDL-Graph-SLAM. The accuracy of the proposed method was an order of magnitude larger than that of LOAM and HDL-Graph-SLAM.
- Published
- 2021