Seabed Sediment Classification using Multifrequency Backscatter

Abstract

Abstract Seabed sediment serves as an important metric for assessing marine ecosystems. Different sectors, including the development and utilization of marine resources, marine engineering, environmental research, marine fisheries, and maritime navigation safety, all depend on the classification and identification of different forms of seabed sediment. Hydrographic surveys employing Multibeam Echosounders provide information that can be used to categorize the types of seabed sediment. The data acquired during the hydrographic surveys using Multibeam Echosounders include bathymetric data and backscatter data. The backscatter data is obtained from the use of Multibeam Echosounder equipment for bathymetric data acquisition. The backscatter data from Multibeam Echosounders can be utilized for determining the classification of seabed sediment as well as for identifying seabed features. This backscatter data represents the types of sediment on the seafloor and can be used for seafloor profiling and acoustic backscatter analysis to determine the several kinds of seafloor sediment. This study uses multi-frequency mosaic backscatter data and an Artificial Neural Network (ANN) approach to classify the different types of seabed sediment in the area of Benoa Port in Bali Province. The research aims to produce a distribution map of sediment types in the study area. A training accuracy of 55% and a testing accuracy of 66% were obtained from the DNN modelling. Three out of six sediment types were identified based on these accuracy results: silty sand, gravelly sand, and sandy silt with gravel corals. According to their distribution, gravely sand had the smallest distribution (5%), while silty sand had the biggest (77%).