Vision Based 3D Pose Estimation for Underwater Vehicle Docking Problem

Abstract

This paper presents an inexpensive, but efficient and accurate vision based pose estimation algorithm for underwater docking problems. While previous docking methods for underwater robots required complicated procedure involving specially designed docking station, the proposed method only needs a simple pattern board as the critical target of the docking task. Our method is based on the estimation of homography induced by projected pattern board into the camera coordinate frame. Each homography is used for estimating the pose of the robot in order to decide the optimal path or sequence of the docking task. The proposed method is validated through real image data set, and the estimated orientation angles were compared against the corresponding IMU data. We have achieved promising results,