Feature-based Underwater Localization using Imaging Sonar in Confined Environments

This paper is focused on AUV operation in confined and unknown environments and introduces a localization and mapping system built on the application of the SURF feature extractor to acoustic images retrieved from an imaging sonar for landmark identification. Adaptations of the feature description and feature matching procedures to the characteristics of acoustic imaging are proposed. A SLAM algorithm following an EKF formulation with augmented state is employed for creating a map resorting to the extracted features and localizing the vehicle within it. The developed solution was subjected to experimental tests based on acoustic data acquired in a water tank structure and under two different scenarios.