Enhancing the Coverage of Underwater Robot Based Mn-crust Survey Area by Using a Multibeam Sonar

Abstract

The authors conducted a field survey of Cobalt-rich Manganese Crusts (Mn-crusts) using a visual 3D mapping system and a multibeam sonar mounted on an underwater robot for studying the distribution of these resources on the seafloor. The multibeam sonar covers an area 4 times that of the visual system and generates bathymetric and backscatter information. This paper shows preliminary results from field trials, showing the correlations between the multibeam and visual mapping data in Mn-crust covered seamounts. Further, the possibility of characterization of the seafloor using multibeam data is investigated by classifying the multibeam data using transfer learning of a classifier trained on visual mapping data. It is advantageous because co-located visual data can be used as a ground truth for multibeam classification. Mn-crust volumetric estimation methods, currently limited to visually mapped regions, can be extended further using these results.