Non-Rigidity Objects Tracking in Dynamic Scenes Using Particle Filter

Abstract

Particle Filter (PF) method is an efficient tool for non-rigidity objects tracking. The paper presents a Bayesian-based PF method for objects tracking in dynamic scenes. The paper discusses the Bayesian estimation algorithm and the PF process. The color histograms are used as the measurement to obtain the optimized posteriori probabilities by comparing the histogram of the particles’ rectangles in the sequence images with the reference histogram. The robust mean technique is applied to ascertain the objects’ positions. The author performed single object and multi-objects tracking in the experiments. In this paper, I compared the PF method with the meanshift algorithm, and the result shows that the PF method is more efficient.