Uncertainty of Abrupt Motion Tracking Using Hidden Markov Model

Abstract Ever increasing the robust tracking of abrupt motion is a challenging task in computer vision due to its large motion uncertainty. visual tracking in dynamic scenarios refers to establishing the correspondences of the object of interest between the successive frames. It is a fundamental research topic in video analysis and has a variety of potential applications like visual surveillance and video analysis. Tracking approach is divided into two categories deterministic and sampling. We have presented a new approach for robust motion tracking in various scenarios. In this paper, we introduceda hidden markov model to solve the local-trap problem and occlusion. Occlusion means when one object is hidden by another object that passes between it and the observer. To estimation of the parameter image object using density grid based normal distribution method is applied. Also, Bayesian filter technique is applied on image object for the purpose of smoothing. In this regard, to reduce the computational cost, less memory, better performance and efficiency. Keyword Abrupt Motion, Bayesian filter, Hidden Markov Model, Point Estimation etc. I. Introduction The robust tracking of abrupt motion is a challenging task in computer vision due to its large motion uncertainty. Visual tracking in dynamic scenarios refers to establishing the correspondences of the object of interest between the successive frames. It is a fundamental research topic in video analysis and has a variety of potential applications, including teleconferencing, gesture recognition, visual surveillance, and motility analysis. Tracking approaches divided into two categories Deterministic and Sampling. At first Deterministic, fast and relatively lower computational cost. The major drawbacks of deterministic for getting trapped in local modes in case of background clutter, distractions, or rapid moving object. The other one, sampling-basedisable to deal with the large motion uncertainty induced by abrupt motions.Earlier works on these lines were proposed by authors isXhuaiuzng Zhou et al hassuggestedto an abrupt motion tracking via intensively adaptive markov-chain Monte Carlo sampling. In this regard, we have astochastic approximation monte Carlo (SAMC) for handling the local-trap problem. In addition, new MCMC sampler for improving sampling efficiency which combines with the SAMC sampling named as Intensively Adaptive -Markov -Chain Monte Carlo (IA-MCMC) sampling. However, SAMC method may cause more computational cost. Reduce the computational cost by introducing a density-grid-based predictive model, Vol: 21, Issue: 2, IEEE Transactions, 2012.This system suffers the major problems are measurement for comparing the tracking accuracy for the target objects with differentsizes.Lower value indicates less local trap problem.Position error includes both mean relativeerror and the standard deviation error. Small value indicates the accurate and stable even usinga small number of samples. To solve this problem we have proposed a Hidden Markov Model on abrupt motio