Tracking underwater cables is an important, real world problem. The vision system described in this paper is designed for use as part of an autonomous underwater vehicle (AUV) solution to this problem.

The system described differs from other vision systems used for this problem and for the related underwater pipe tracing problem, in that it incorporates the novel and interesting idea of using bidimensional histograms to segment real-time video images, before converting them into contour images and detecting the cable.

Pragmatic information is used to make the system fast enough, in that cables go from image top to bottom as the AUV is tracking the cable. It is also used to make the system robust enough because there is recovery when the cable gets buried in sand, or otherwise lost by the tracker. The system has been tested on a videotape of several tracking sequences with a remote underwater vehicle (RUV) driven from the surface, but not on a real AUV.

This use of pragmatic information offers a high quality and complete solution to the cable tracking problem; we can expect AUVs to replace surface controlled RUVs, not only for the simpler pipe following problem, but also for this problem.