Fig. 1. Coordinate system conversion

Fig. 2. System integration block diagram

Fig. 3. Comparison of different detectors. (a) (c) Proposed characteristic detector; (b) (d) feature detector based on Shi Tomasi and BRISK

Fig. 4. Lidar odometer point cloud scan. (a) Lidar odometer scan front view; (b) lidar odometer scan top view

Fig. 5. Scan context descriptor

Fig. 6. KITTI 05 and KITTI 00 global and local point cloud maps. (a) KITTI 05 global point cloud map; (b) KITTI 05 local point cloud map; (c) KITTI 00 global point cloud map; (d) KITTI 00 local point cloud map

Fig. 7. Factor diagram structure

Fig. 8. Odometer gravity vector drift comparison of proposed algorithm and V-LOAM. (a)(b)(c) Localization effect of proposed algorithm when vector drift occurs in V-LOAM; (d)(e)(f) gravity vector drift phenomenon that occurs in V-LOAM

Fig. 9. Odometer positional drift comparison of proposed algorithm and V-LOAM. (a)(b) Positioning effect of proposed algorithm when V-LOAM has positional drift; (c) (d) positional drift phenomenon in V-LOAM

Fig. 10. Local map accuracy comparison of proposed algorithm and V-LOAM. (a) Local map effect of proposed algorithm; (b) V-LOAM local map loopback failure

Fig. 11. Comparison of proposed algorithm, LeGO-LOAM, V-LOAM in KITTI 05, 06, 09 with true value trajectory. (a) KITTI 05 track comparison; (b) KITTI 06 track comparison; (c) KITTI 09 track comparison; (d) KITTI 05 track details; (e) KITTI 06 track details; (f) KITTI 09 track details;

Fig. 12. Comparison of proposed algorithm, LeGO-LOAM, V-LOAM in KITTI 00 with true value trajectory. (a) KITTI 00 track comparison; (b) KITTI 00 track details