Pose Estimation Considering an Uncertainty Model of Stereo Vision for In-Water Ship Hull Inspection

Abstract

For visual inspection of underwater structures by using an autonomous underwater vehicle (AUV), the relative pose between the AUV and the underwater structure needs to be estimated accurately. In this paper, a vision based pose estimation method for in-water ship hull inspection using 3D point cloud data provided by a stereo vision system is proposed. In particular, a systematic procedure to evaluate the uncertainty of feature point location on the obtained stereo images is formulated and applied. An experimental result is presented to demonstrate the practical feasibility of the proposed approach.