Fig. 1. Flow chart of point cloud clutter filtering algorithm

Fig. 2. Introduction of lidar clutters in sailing scene

Fig. 3. Schematic diagram of position and world coordinate system of vehicle body sensor

Fig. 4. Principle diagrams of point cloud generated by lidar,and panoramic image generated by panorama camera. (a) Principle diagram of point cloud generated by lidar; (b) panoramic image generated by panorama camera

Fig. 5. Comparison of data volume between panoramic images and point cloud during 0.1 s. (a)-(c) Panoramic images taken in less than 0.1 s; (d) diagram of two-dimensional point cloud in 0.1 s

Fig. 6. Network diagram of Darknet-53

Fig. 7. Mapping relationship of panoramic image and world coordinate system. (a) Top view; (b) side view

Fig. 8. Relationship between azimuth of panoramic image and pixels of horizontal axis of image (top view)

Fig. 9. Schematic diagram of influence of water surface height on vertical axis pixels

Fig. 10. Flow chart of adaptive matching algorithm

Fig. 11. Registration diagrams of target point clouds with different vertical axis pixel values of image correspond to the 0° pitch angle

Fig. 12. Point cloud clutter removal effect of fusion of lidar and panorama camera. (a) Three-dimensional point cloud diagram; (b) two-dimensional point cloud diagram; (c) point cloud pixel image; (d) panoramic image; (e) target detection for panoramic image; (f) point cloud mapping of target ship

Fig. 13. Clustering effect of lidar point cloud before and after clutter filtering. (a) Three-dimensional point cloud diagram; (b) three-dimensional point cloud clustering diagram; (c) three-dimensional point cloud diagram after clutter filtering; (d) three-dimensional point cloud clustering diagram after clutter filtering

Fig. 14. Changes in ships with point cloud contour in sailing.（a）3D point cloud distribution before target ship encounters local ship；（b）3D point cloud distribution in the encounter between target ship and ship；（c）3D point cloud distribution after target ship encounters local ship；（d）3D point cloud distribution map after target ship leaves ship