The application described here concerns the detection and the real-time tracking of submarine pipelines from a sequence of underwater images acquired from TV-cameras mounted on a remote operated vehicle (ROV). The problem involves both perception and robotics, in that sensory-motion data are to be acquired and dynamically used for subsequent actions by the system. The basic processing concerns the extraction of the pipe contours which are returned as two straight lines. Motion effects on acquired images have been reduced using a Kalman filter for integrating multiple measurements over time. The filter, positioned at the output of the image processing system improves the precision in the computation of the straight line equations with a little overload. This work is originated from the interest of Snamprogetti in enhancing the level of automation in submarine pipeline inspection.