Underwater Robot Localization using Artificial Visual Landmarks

Abstract

Accurate localization for underwater robot is a big challenge in the mobile robot community. We will use acoustic and visual-based approaches to solve the problem. In this paper, we focuses on the artificial landmark visual-based localization. The robot carries the camera to localize its position by calculating the camera's viewpoint through looking at the landmarks. The position of landmarks is known in a world-centered coordinate system (WCCS). A method is presented in this paper to recover the camera's viewpoint with minimum three feature points and a single image from one camera. Two steps are used in this paper: first step is to calculate the feature points' 3D coordinates in a camera centered coordinate system (CCCS); second step is to obtain a closed-form solution through the geometric transformations to map the 3D points from CCCS to WCCS. The algorithm is robust and efficient. It uses the fewest feature points required so far to deal with the same problem.