A Novel Sub-Bottom Profiler Seabed Sediment Classification Method Based on BPNN With Biot-Stoll Model and Attenuation-Based Model

Abstract

The exploitation and utilization of seabed sediments provide vital significance in many fields. Recently, t he classification of seabed sediments using sub-bottom profiler(SBP) data has become a research focus. Specifically, SBP data can be applied not only for recognizing individual stratigraphic layers but also for classifying seabed sediments by inversion models. To improve the sorting effect of gravel and mud simultaneously, we propose a sediment classification method based on aback propagation neural network (BPNN) with the Biot-Stoll model and an attenuation-based model. In this method, two datasets of the mean grain size derived from these two models are trained through the BPNN classifier to classify seabed sediments. The proposed method is verified through SBP data and in-situ sampling data collected from the sea north of the Shandong Peninsula, China. The experimental results show that the overall accuracy of sediment classification is 89.4%, and the classification accuracy of gravel and mud reach 91.4% and 93.3%, respectively, confirming that gravel and mud can be more accurately distinguished based on the proposed method than the single Biot-Stoll model and the single attenuation-based model.