다목적 인공 표식물을 이용한 수중로봇의 자율도킹을 위한 영상기반의 프레임워크

Abstract

We propose an efficient and accurate vision based framework for autonomous docking tasks of underwater robots. We designed a self-similar landmark (SSL) based passive docking platform which could be flexibly used for both landmark detection and the pose estimation steps. That is, in the homing phase, the robot is guided by SSL target until the target comes close to the camera. When the AUV reaches to a vicinity position of the target, the pose estimation process comes into play. The proposed method is validated through real image data sets, and we have achieved promising results.tion and the pose estimation steps. That is, in the homing phase, the robot is guided by SSL target until the target comes close to the camera. When the AUV reaches to a vicinity position of the target, the pose estimation process comes into play. The proposed method is validated through real image data sets, and we have achieved promising results.