基于模型的水下机器人视觉悬停定位技术

Model-based vision for station keeping of an underwater vehicle

根据7000米载人潜水器水下作业的需要和特点,提出了一种基于单目视觉位姿测量原理的水下机器人悬停定位方法,构建了包含水下目标识别与跟踪、三维位姿计算及水下机器人控制三项关键技术的水下演示实验系统.以自行研制的水下机器人控制系统实验研究平台为实验载体,以4个按正四面体分布的小球构成的合作目标为观察目标,在室内实验水池中实现了水下机器人悬停与定位演示实验.实验结果显示,系统能够引导机器人运动到指定位姿并实现稳定的悬停和定位.该项研究为基于模型的单目视觉三维位姿测量技术开辟了水下机器人应用新领域.

A monocular vision-based station keeping method is presented for the working demand of the 7000-meter manned underwater vehicle. The experimental system consists of three key techniques： underwater target recognition and tracking, position and orientation estimation, and underwater vehicle control. Accurate station keeping is demonstrated in experiments with our four-DOF underwater vehicle and the observed target which is a regular tetrahedron with four ball-like culminations in our indoor water tank. The experimental results show the vehicle can not only automatically move to a desired station but also maintain a fixed position and orientation guided by the system. Our research extends a new field of underwater vehicle for the 3D pose measurement technique of model-based monocular vision.