Fig. 1. Framework of the SLI-SLAM method

Fig. 2. Cluster analysis based on the flatness degree

Fig. 3. Results of point cloud segmentation and feature extraction. (a) Original point cloud; (b) ground point cloud; (c) split point cloud; (d) edge feature point; (e) plane feature point

Fig. 4. D3p-S network structure

Fig. 5. Point cloud semantic segmentation. (a) Original point cloud; (b) semantic segmentation point cloud

Fig. 6. Dynamic obstacle removal effect. (a) Before dynamic object removal; (b) after dynamic object removal

Fig. 7. Factor graph optimization

Fig. 8. Experimental platform

Fig. 9. Comparison of the semantic segmentation results. (a) (b) Original image; (c) (d) image segmentation effect of the Deeplabv3+; (e) (f) image segmentation effect of the D3p-S

Fig. 10. Movement trajectory and APE error distribution of the autonomous localization test. (a) (c) Campus environment;(b) (d) KITTI-00 sequence

Fig. 11. Environmental mapping results. (a) SLI-SLAM; (b) LeGo-LOAM; (c) LIO-SAM

Fig. 12. Urban road environment mapping results. (a) (b) (c) Camera image; (d) (e) (f) semantic segmentation of the images;(g) (h) (i) semantic point cloud map; (j) semantic map of the urban road environment construction