Fig. 1. Flowchart of fusion-based vehicle speed estimation technique

Fig. 2. Differentiating foreground from background using ground points

Fig. 3. Preliminary segmentation of main point cloud and noise point cloud in overhead 2D view and selection of seed point regions

Fig. 4. Restoration to 3D point cloud and selection of clustering seed points based on centroids

Fig. 5. Traffic conditions in various scenarios

Fig. 6. Visualization of calibration results in some scenarios

Fig. 7. Speed measurement results of descending after straight driving from near to far

Fig. 8. Speed measurement results of downhill and straight driving from far to near

Fig. 9. Results of vehicles traveling from far to near and turning at a gentle corner at the end

Fig. 10. Speed measurement results of vehicles approaching a stop state

Fig. 11. Speed measurement results of vehicles during turning

Fig. 12. Image-prioritized multi-target vehicle extraction and localization results