A Line/Plane Feature-based Lidar Inertial Odometry and Mapping

We propose a line/plane feature-based lidar inertial odometry and mapping method, for real-time 6-DOF pose estimation with unmanned ground vehicle. The accuracy of the estimated pose is mainly due to the correctness of the corresponding points matching. It is difficult with Velodyne-type sensors due to the sparse and non-uniform point clouds that they produce. In this paper, we describe a feature-based approach using segmentation and clustering algorithm, which results in mathematically principled line and plane features. A Levenberg-Marquardt optimization method then uses the line and plane features to solve different components of the six degree-of-freedom transformation across consecutive scans. We also use the pose-graph to perform loop-closure detection. We evaluate the performance of our algorithm by using datasets gathered from building environments with unmanned ground vehicles, and we compare our results against the state-of-the-art method LOAM. The results indicate that our method can achieve better accuracy.