MSPCC: A Shipborne LiDAR Data Hierarchical Compression Approach in Dynamic Port Environment

The application of shipborne light laser detection and ranging (LiDAR) has significantly enhanced the autonomous navigation capabilities of maritime autonomous surface ships (MASSs) and provided a new solution for shore-based monitoring of ship dynamics. However, in port environments, the large volume and high sparsity of point cloud pose a major challenge to data transmission and storage, hindering efficient ship-to-shore information exchange and remote control. This article proposes a hierarchical compression approach for navigation environment point cloud, combining deep learning with the portable network graphics (PNG) format. The approach first applies motion compensation to the measured data using the shipborne inertial measurement unit (IMU), then transforms the unstructured environmental point cloud into a structured matrix, and finally performs hierarchical data compression on the feature images. Real-ship test results demonstrate that the reconstructed LiDAR data exhibit high quality and geometric fidelity, with the root mean square error (RMSE) as 0.148 m and the symmetric nearest neighbor root mean squared error (SNNRMSE) as 0.1191 m. The peak signal noise ratio (PSNR) shows an average improvement of approximately 11.5%, while the average encoding time is only 150 ms. This approach thus improves the compression efficiency of perception data during ship-shore interaction, while enhancing safety for MASS navigating in port waters.