Seabed Sediment Classification Using Spatial Statistical Characteristics

Conventional sediment classification methods based on Multibeam Echo System (MBES) data have low accuracy since the correlation between features and has not been fully considered. Moreover, their poor resistance to residual error of MBES backscatter strength (BS) processing also degrades performances. Toward these problems, we propose a seabed method using spatial statistical extracted from angular response curve (ARC), topography, geomorphology. First, reduce interference noise beam pattern correction, robust combining Generic Seafloor Acoustic Backscatter (GSAB) model Huber loss function estimate parameters ARC which is strongly correlated with sediments. Second, feature set constructed by AR composed GSAB parameters, BS mosaic its derivatives, topography derivatives characterize After that, selection probability map acquisition are employed random forest algorithm (RF). Finally, denoising final generation proposed applied maps obtain reasonable distribution clear boundaries classes. We implement experiments achieve 93.3%, verifies validity our method.