Visual tracking system for a mobile robot using colour histograms

Abstract

This work describes the visual system of a mobile robot based on a pan-tilt structure which has been endowed with the ability of tracking moving object using merely color information. Moving objects in the field of view of the camera are detected and the color feature of the most relevant regions is selected as the pattern to follow. Color histograms are used as reliable descriptors to model the appearance of objects. In order to handle with illumination changes a simple adaptation scheme is used. Results show that this system is reliable and fast enough to perform real time tracking of a moving object.