Towards stereo vision SLAM based pose estimation for ship hull inspection

摘要

The introduction of foreign species imposes one of the greatest threats to any indigenous ecosystem. As the world's main means of transport of global goods, ships introduce the major part of unwanted marine organisms of which a significant part resides on the hull of ships. Before taking measures to clean these invading organisms, ship hull inspections are necessary. To this end, using autonomous underwater vehicles (AUV) is more desirable than sending divers, which has inherent problems in cost. For the AUV to operate, a pose (position and orientation) estimation system is required. This paper reports on the current development towards a real time, stereo camera and SLAM (Simultaneous Localization and Mapping) based localization system for underwater ship hull inspection. Based on underwater footage filmed at ship hull inspections, the development of the first-generation extended Kalman filter (EKF) SLAM system using a stereo camera and the employed landmark tracking techniques are presented. Serving as a basis for future development, an evaluation of the system's ship hull tracking capability, measurement precision and real-time performance is given by experiments. Preliminary results show that this on-board sensor based system has the potential to be a promising instrument for ship relative pose estimation.