Classification of coastal zone point clouds using multi-channel fusion of UAV-borne photon-counting LiDAR data

Coastal zones are densely populated and subject to frequent human activities, making accurate surveying essential for ecological monitoring and navigation safety. Compared with full-waveform LiDAR, photon-counting LiDAR enables more efficient data acquisition and denser point clouds, but also introduces substantial noise that complicates point cloud classification. Using coastal data collected by a four-channel UAV-borne photon-counting LiDAR system, this study proposes a supervised 3D point cloud classification algorithm. The method employs multi-dimensional grid-based neighborhoods and integrates multi-channel features, including polarization characteristics, to directly classify noise, land, sea surface, and seabed without prior denoising. In addition, a category-adaptive optimization algorithm is introduced to correct misclassified points. Experimental results show that the proposed method achieves an overall accuracy of 96% and a Kappa coefficient of 0.92. The classified point clouds further support integrated marine-terrestrial surveying applications.