Uncertainty Calibrated Markov Chain Monte Carlo Sampler for Visual Tracking Based on Multi-shape Posterior

Abstract

We present a novel tracking system that adaptively selects a shape of the posterior over time, where the selection is efficiently performed by the uncertainty calibrated Markov Chain Monte Carlo (UCMCMC) sampler. In conventional trackers, the posterior is typically described by a single prior distribution. On the other hand, our tracker allows the posterior to have multiple prior distributions, namely normal and Student’s t distribution, and to choose an appropriate distribution according to the tracking environment. The optimal distribution is determined by the UCMCMC sampler in the process of reducing the uncertainty of the shape. Experimental results demonstrate that the proposed multi-shape posterior helps improve the tracking performance in terms of accuracy.