Applying Deep Learning to Bathymetric Lidar Point-Cloud Data for Classifying Submerged Environments

Sub-sea environments host rich biodiversity and resources, playing a crucial role in climate regulation while supporting human activities such as fishing, transport, and resource extraction. To maintain healthy underwater ecosystems and ensure sustainable operations, accurate mapping and monitoring are essential. Among available technologies, airborne LiDAR bathymetry (ALB) stands out for its ability to capture detailed subsea data, but handling the enormous datasets it generates remains a major challenge. In this study, we propose a novel preprocessing pipeline combined with deep learning models—Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM)—to classify submerged environments using data from Fjøløy, Stavanger, Norway. LSTM achieved a higher classification accuracy (95.22%) compared to BiLSTM (94.85%). To further assess robustness, classification reliability was evaluated through confidence scores, offering insights into model dependability. The results demonstrate the potential of deep learning for ALB classification and provide practical value for mapping authorities and practitioners in producing reliable seabed maps and nautical charts.