Visual tracking with double-layer particle filter

Abstract

Particle Filter is one of the most widely used algorithm in object tracking, because it can handle the nonlinear and/or non-Gaussian problems. However due to loss of diversity among particles, its tracking performance is not ideal. In order to solve this problem, in this paper, a novel double-layer particle filter is proposed. The particles are divided into two layers: the parent particles and the child particles. The child particles are used to remember the latest state of the parent particles and optimize the parent particles. In addition, only the parent particles are updated during re-sampling while the child particles remain unchanged, which maintains the diversity of the particles to some extent. Finally, the parent particles are used to estimate the state of the object. Experimental results show that the tracking performance of the proposed double-layer particle filter outperforms that of the basic particle filter.