Seafloor Sediment Classification using BP Neural Network

Seafloor sediment classification plays an important role in the development and utilization of marine resources, marine engineering, underwater communication, marine military and so on. In this paper, a BP-D7 network model is proposed based on the seafloor echo data of single beam sonar. This method can automatically extract sonar signal features and classify seabed geological types. Compared with machine learning algorithm, KNN and decision tree, the classification performance is better. Regarding the limited number of samples, a data expansion method of equal interval sampling is proposed and the classification effect is improved. The experimental results show that the proposed seafloor classification model is competitive and effective.