Fig. 1. Schematic diagram of dynamic line-scan point cloud correction method based on two-dimensional image reference

Fig. 2. Process of point cloud profile generation and texture mapping

Fig. 3. Deviation of reprojection image caused by motion perturbance

Fig. 4. Process of preliminary correction based on global transformation of reprojection image

Fig. 5. Process of fine correction based on optical flow analysis of reprojection image

Fig. 6. Influence of point cloud profile pose perturbance in different degrees of freedom on the reprojection image

Fig. 7. Process of correction compensation based on low-noise reference

Fig. 8. System diagram of measurement experiment

Fig. 9. Rail sample scanning scene and measured point cloud. (a) Measurement scene and rail sample; (b) line-scan point cloud

Fig. 10. Rail sample point cloud correction results of measurement with large perturbance. (a) Point clouds with mapped textures; (b) point clouds reflecting shape characteristics

Fig. 11. Quantitative accuracy analysis of measurement with large perturbance

Fig. 12. Rail sample point cloud correction results of measurement with small perturbance. (a) Point clouds with mapped textures; (b) point clouds reflecting shape characteristics

Fig. 13. Quantitative accuracy analysis of measurement with small perturbance

Fig. 14. Point cloud correction results of using the proposed method. (a) Point clouds with mapped textures; (b) point clouds reflecting shape characteristics

Fig. 15. Quantitative accuracy analysis of the correction effect using the proposed method

Fig. 16. Comparative experiment of point cloud stitching constraint using pose estimation by attached sensors. (a) System diagram; (b) point cloud with mapped textures and point cloud reflecting shape characteristic; (c) quantitative accuracy analysis

Fig. 17. Results of ablation experiment. (a) Without the first initial correction based on global transformation of reprojection image; (b) without the second fine correction based on optical flow analysis of reprojection image; (c) without the third correction compensation based on low-noise reference; (d) contrast point cloud of the complete correction process