Tracking Appearances with Occlusions

Ying Wu, Ting Yu and Gang Hua
{yingwu, tingyu, ganghua}@ece.northwestern.edu
Computer Vision and Pattern Recognition (CVPR’03), Madison, WI, June, 2003

Abstract:

Occlusion is a difficult problem for appearance-based target tracking, especially when we need to track multiple targets simultaneously. To cope with the occlusion problem explicitly, this paper proposes a dynamic Bayesian network as a generative model by accommodating an extra hidden process of occlusion in the probabilistic framework to stipulate the conditions on which the image observation likelihood is calculated. The statistical inference of such a hidden process can reveal the occlusion relations among different targets, which makes the tracker more robust against partial even complete occlusions. In addition, considering target appearances are affected by views, another generative model for multiple view representation is proposed by adding a switching variable to select from different view templates. The integration of the occlusion model and multiple view model results in a complex dynamic Bayesian network with hidden processes for the switch of targets' templates, the states of targets' motion, and the occlusions among different targets. The tracking and inferencing algorithms are implemented by the sampling-based sequential Monte Carlo strategies. Our experiments show the effectiveness of the proposed probabilistic models and the algorithms.

Results:

Publication:

Ying Wu, Ting Yu and Gang Hua, “Tracking Appearances with Occlusions”, in Proc. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR’03), Vol I, pp.789-795, Madison, WI, June, 2003.