Lidar Odometry and Mapping Based on Two-stage Feature Extraction

We propose a lidar odometry and mapping method based on two-stage feature extraction for real-time 6-DoF pose estimation of mobile robots. Firstly, we take a preprocessing operation to filter out point cloud noise and downsample it. Then, in lidar odometry step, we extracts edge and plane features in a stage-wise manner according to different constraints to obtain more robust features. During the mapping step, we add time constraint and distance constraint to key-frame-based loop-closure detection, which greatly reduce the pose estimation error caused by drift. We validated our method on datasets collected in both indoor and outdoor environments. We compared our method with the classic LOAM algorithm. The experimental results show that our method achieves better positioning accuracy and mapping precision.