Autonomous target tracking by underwater robots based on vision

This paper proposes and demonstrates the performance of a target tracking system designed for underwater robots (UR) based on the image data of the target, captured by a single CCD camera mounted on the robot. The use of target features for the derivation of navigational commands for the UR is addressed. The proposed system simplifies the underwater navigation problem by introducing additional parameters using visual image. Underwater positioning is one of the major problems encountered by URs. In the methodology proposed, such positioning is not required making the navigation problem easier. Here, the behavior of the target object in the vicinity of the UR is observed using CCD data and is interpreted into navigational commands. Implementations of this system for underwater cable tracking and moving object following missions using the UR test-bed available at the University of Tokyo are presented.