In this article, we are concerned with tracking an object of interest invideo stream. We propose an algorithm that is robust against occlusion, thepresence of confusing colors, abrupt changes in the object feature space andchanges in object size. We develop the algorithm within a Bayesian modelingframework. The state space model is used for capturing the temporal correlationin the sequence of frame images by modeling the underlying dynamics of thetracking system. The Bayesian model averaging (BMA) strategy is proposed forfusing multi-clue information in the observations. Any number of objectfeatures are allowed to be involved in the proposed framework. Every featurerepresents one source of information to be fused and is associated with anobservation model. The state inference is performed by employing the particlefilter methods. In comparison with related approaches, the BMA based tracker isshown to have robustness, expressivity, and comprehensibility.
展开▼
