Tracking and Recognition: A Unified Approach on Tracking and Recognition

Abstract

This paper proposes a unified approach on tracking and recognition .Object tracking is done at low level and recognition is done at high level. Traditional tracking methods give importance to low level image correspondences between frames. High level image correspondences are used for reliable tracking. Online and Offline models are used for both tracking and recognition which is done simultaneously. Thus high level offline model is combined with low level online model to increase the tracking performance. Onine model used for tracking is given to the video based recognition and at same time offline model plays important role to recognize the category of the object. This method is useful to handle difficult scenarios like abrupt change, background clutter, pose variations, occlusion and morphable objects. This is based on study of different IEEE papers.