Effective Motion Tracking of Moving Persons/Objects Using MCMC Sampling Method

Abstract

The robust tracking of the abrupt motion is a challenging task in the recent field of computer vision. For visual tracking various tracking methods such as particle filters and by using Markov-Chain Monte Carlo (MCMC) method have been proposed , but these methods grieve from the local-trap problem and abrupt motion un certainity. In this paper, we introduce the Stochastic Approximation Monte Carlo (SAMC) sampling method into the Bayesian filter tracking framework for handling the local-trap problem. In addition for improving the sampling efficiency, and propose a new MCMC sampler with intensive adaptation. This is done by combining the SAMC sampling with a density-grid-based predictive model. The proposed method is very effective and computationally efficient in addressing the abrupt motion problem. The proposed method is named as Intensively Adaptive Markov Chain Monte Carlo (IA-MCMC).The experiment results for various videos with single and multiple objects have been proposed.