Semantic Mapping on Underwater Environment Using Sonar Data

Abstract

The use of robots in underwater exploration is increasing in the last years. The automation of the monitoring, inspection and underwater maintenance tasks requires the understanding of the environment. One of the key issues of these systems is to recognize in the objects in the scene. This paper proposes a method to provide a semantic mapping using acoustic images acquired by forward looking sonar (FLS). The method represents the environment using Gaussian probability density functions. Furthermore, we efficiently segment and classify the structures in the scene. Finally, we create a semantic map of the scene. We evaluate the method in a real dataset acquired by an underwater vehicle performing autonomous navigation and mapping tasks in a harbor area.