Fig. 1. Module diagram of proposed method

Fig. 2. Fan－shaped grid model

Fig. 3. Target segmentation effect diagram

Fig. 4. Schematic diagram of point selection in dynamic region

Fig. 5. Schematic diagram of corresponding point query

Fig. 6. MF－RANSAC registration diagram. (a) m－5 frame; (b) m frame; (c)(d) optimal matching results of multiple feature point using static obstacle and dynamic obstacle

Fig. 7. Schematic diagram of fan－shaped grid merge

Fig. 8. Similarity calculation of feature matrix

Fig. 9. Collection platform for KITTI dataset

Fig. 10. Real vehicle dataset. (a) Collection platform; (b) satellite map of the experimental section

Fig. 11. Precision－recall curve under KITTI dataset. (a) KITTI 00; (b) KITTI 02; (c) KITTI 05; (d) KITTI 08

Fig. 12. Precision－recall curves under real vehicle dataset

Fig. 13. Error comparison chart of each algorithm. (a) Distance error; (b) yaw error

Fig. 14. Leaving scene M1. (a) Dynamic target tracking result; (b) ScanContext feature map; (c) SR－Context feature map

Fig. 15. Returned scene M2. (a) Dynamic target tracking result; (b) ScanContext feature map; (c) SR－Context feature map