A vision system for underwater real-time control tasks

The framework of the research described in this paper is artificial object recognition in an unstructured environment. The mixture of natural and artificial objects as a general scenario has been an ill-defined problem up to now. One of the most interesting and challenging environments where the mentioned mixture appears is the sea-bed. This paper describes in-progress results in the development of a computer vision system capable of automatically detecting and tracking a submarine cable, using vision as the main source of information and producing results in real time.