An Unsupervised Learning Approach for Seafloor Mapping Using Underwater Robot

Abstract

Seafloor mapping has great advantages in natural disaster preparedness as well as seafloor resource explorations and operations. However, it is limited in not having sufficient seafloor mapping data set. The paper introduced an unsupervised learning-based seafloor classification method for seafloor mapping. The team successfully classified the seafloors using autoencoder and k-means clustering algorithm and conducted tests using the image data obtained from the actual seafloors.