Object-tracking based on particle filter using particle swarm optimization with density estimation

Visual object-tracking is a fundamental task applied in many applications of computer vision. Many different tracking algorithms have been used ranging from point-tracking, to kernel-tracking, to silhouette-tracking based on different appearance models chosen. This paper investigates the particle filter that is used as a tracking algorithm based on the Bayesian tracking framework. The problems that the particle filter tracking technique suffers from are degeneracy and the impoverishment degradation. These two issues are addressed by the use of Particle Swarm Optimization (PSO) as the sampling mechanism. In particular, particles are generated via the PSO process in order to estimate the importance distribution. Two density estimation methods are used, one is a parametric method using the Half-Normal distribution fitting, and the other is a non-parametric method using kernel density estimation. The experiments revealed that the non-parametric density estimation method combined with PSO outperforms the other comparison algorithms.