High-precision underwater navigation using model-referenced pose estimation with monocular vision

Abstract

In this study, model-referenced pose estimation using monocular vision is applied to the navigation of an underwater vehicle in an underwater environment. The relative spatial information between the vehicle and the nearby underwater structure is obtained by employing 3D model-referenced tracking techniques. The obtained relative pose enables the underwater robot to localize around the underwater structure whose geometric model is known a priori. To demonstrate the performance of the proposed algorithm a set of experiments was carried out in a test tank and its results are shown.