Implementation of active vision on mobile robot Ben Hope: a real-time edge-tracking system

Abstract

This thesis describes the successful development of an active vision module, the edge-tracking vision system, on a transputer-based system which was prototyped on a SUN 3/140 workstation using a MATROX frame store. The transputer system consists of a root TRAM, a video TRAM and a graphic TRAM which will eventually be ported to the mobile robot Ben Hope for autonomous navigation. Due to the huge amount of processing required in active vision, computing in parallel is essential.

The problem of unpredicted environmental disturbances can cause undesirable results using the theory of motion of parallax. Some solutions to these problems are discussed in the thesis.

The system was simulated in real-time, performing reasonably well at different location within the laboratory environment without having to change previously specified parameters for proper functioning of the vision system.