An Improved Color-Based Particle Filter for Object Tracking

The object tracking problem in a nonlinear and/or non-Gaussian circumstance can be solved by particle filter estimation based on the concept of sequential importance sampling and the use of Bayesian theory. An improved object tracking scheme is proposed, which is based on the Markov chain Monte Carlo (MCMC) particle filter and object color distribution. This scheme is robust to clutter, deformation of non-rigid object, rotation, and partial occlusion. In our scheme, a novel MCMC sampling method is applied to adjust particles distribution to obtain a more efficient particle filter. Object appearance is represented by color distribution within the specified primitive region. Experiments results demonstrate that the improved particle filtering method is robust and effective for object tracking.