A visual tracking system for a biomimetic autonomous underwater vehicle

Abstract

A visual tracking system for the guidance and navigation of a biomimetic-autonomous underwater vehicle (BAUV) is proposed. Video camera provides precise guiding when the BAUV is close to its target. We design the guidance system that utilizes vision sensor to approach a target. The purpose of the visual sensor is to assist the position estimation of the vehicle and the environment. BAUV swims with cyclic oscillations. Hough transform is applied for resolving the swinging features on the image plane due to the yawing, rolling and pitching of BAUV. A kinematic model is derived to describe the motion of BAUV, and the stereo imaging is applied to estimate relative distance between the BAUV and the underwater target. An extended Kalman filter that combines observational information with navigational data is shown to be effective in reducing positional uncertainty of the BAUV. Numerical simulations and experimental results in a water tank show the validity of the proposed method