A hierarchical learning framework for seafloor scene classification

Abstract

Present study consists an experimental study of the framework for seafloor scene classification with AUV data. In this framework, the dictionary is dynamically learned with new features, which are extracted from marine images. Sparse representation is further fed into a Support Vector Machine for scene classification. Extensive experimental results on the AUV data of southeast coast of Tasmania have shown that the proposed framework provides a satisfying approach with respect to the seafloor scene classification.