Mono visual odometry for Autonomous Underwater Vehicles navigation

This paper presents a mono visual odometry algorithm tailored for AUVs navigation. The main goal of the paper is to test different odometry solutions to find out which one is the most suitable for AUVs navigation. Two different features detector, SIFT and SURF, to compute the vehicle motion are compared. The algorithms were tested with a dataset acquired in September 2018 by the compact Zeno AUV in the Haifa bay, Israel, and the results were compared with the path followed by the vehicle, estimated exploiting GPS, IMU and DVL, that are more accurate than the camera. The achieved results are promising; the proposed strategy could support and enhance DVL-aided navigation or could represent an alternative, for instance, in DVL-denied scenarios or for low-cost AUVs (without a DVL on board).