A vision system for pose estimation of an underwater robot

摘要

This paper presents a vision system designed for an underwater robot to facilitate measuring the distance to an inspected surface and its orientation with respect to the robot. The system uses a set of laser modules to project a rectangular dot pattern on the inspected surface. The vision system is used to measure the dots' locations in the 3D space and calculate the distance to the surface, together with the pitch angle and the yaw angle of the camera. A novel calibration procedure is presented, which allows for constructing high resolution camera-laser systems, without requiring a perfect alignment of laser beams and camera optical axis. An underwater vision system simulation environment, based on open-source software, is also introduced and extensively used to test the developed algorithms. The experimental results, obtained in a laboratory tank and a pool, confirm high performance of the system.