Graphical Model based Cue Integration Strategy for Head Tracking

Abstract

To achieve robust system, more and more vision researchers take into account fusing multiple visual cues. In this paper, we propose a novel strategy to integrate multiple naive cues for head tracking. Firstly, a cue dependency model is constructed via graphical model. Secondly, a new inference procedure based on non-parametric belief propagation is built for cue integration. The work presented is thus a general framework easy to extend for other computer vision research problems. Experimental results imply that the strategy we propose is effective, and it is robust without estimation of cue reliability.

Perceptual stimulus — A Bayesian-based integration of multi-visual-cue approach and its application

“…More details of setting the parameters can be found in ref. [33]. The experimental results show that the LogOP-based tracker is sensitive to the noise because multiplication will amplify the noise, while the LOP-based tracker can tolerate noise to some extent, but it decreases the accuracy of the estimation.…”

“…However, Monte Carlo computing in these methods depends heavily on sampling positions. To compensate for the insufficiency caused by the absence of information from belief of nodes, we proposed a stratified sampling belief propagation algorithm to estimate the belief of each node [32,33] , and applied the algorithm successfully in tracking multiple visual targets.…”

“…For example, the discriminability of the color cue decreases quicker than that of the contour cue under weak illumination. To address this problem, we propose a PGM to describe the correlations among cues, and use a sequential mean field algorithm to make inference [33] on the PGM. The potential functions are defined as an asymmetric mixture Gaussian [43] .…”

Multi-Camera Tracking with Adaptive Resource Allocation

“…However, the relations are subjective and hard to be generalized, and there is no discussion of their performance. The graphical model employed in Zhong et al (2006) might not be used in system with many sensors due to its complexity.…”

“…The co-inference of tracker state based on shape and color is optimized jointly in Wu and Huang (2001). Cue dependency is defined with a graphical model and cue integration is performed by Bayesian inference in Sherrah and Gong (2001), Zhong et al (2006). Sherrah and Gong (2001) proposed heuristics to estimate the reliability of each modality based on relation between modalities.…”