Design and implementation of a camera based underwater localization module for autonomous navigation

The increasing demand for easier, faster and safer ways to intervene in large underwater environments has grown over the past years. By developing a low cost fully autonomous underwater vehicle (AUV) that will meet today's needs of autonomy, reliability and versatility, uWare Technologies has taken a step forward to solve these issues. One of the most critical aspects of an AUV is the underwater localization system. The challenges of the marine environment incite to develop innovative technologies. This master thesis was carried out in uWare and aimed to implement and evaluate a camera-based SLAM algorithm to be used for autonomous navigation purposes. We start by reviewing the most important concepts of Computer Vision for pose estimation. Then, we present the SLAM approach to solving the localization and mapping problem. Multiple SLAM implementations are reviewed, and we select the one that suits the most the requirement of uWare. The selected algorithm is then analyzed in details and implemented in the AUV. After evaluating the algorithm in different scenarios, some potential avenues for improvement are explored.