Abstract Architecture of an Object-based Tracking System Using Colour Segmentation

email: Rainer.Bischoff�UniBw-Muenchen.de This paper presents a robust specialised architecture for real time tracking, using colour segmentation. Most of the existing object-tracking algorithms extract the features which constitute the object (usually called tokens), and track them from one frame to another. We propose a new architecture to solve the tracking problem at object level. The first step is the extraction of the object from the scene. The proposed architecture uses the discriminatory properties of three colour attributes: Hue, Saturation and Intensity, in order to segment the object. The second step consists of the computation of the centroid of the object by using the enhanced image from the segmentation module. As a last step, tracking of this centroid is achieved in a sequence of images. The processor presented here can provide the sequence of centroids at video rate