Research of Kernel Particle Filtering Target Tracking Algorithm Based on Multi-feature Fusion

摘要

The standard particle filter usually fails in the scene of small system noise or weak dynamic models. The robustness of majority tracking algorithm is not high due to using only a single target feature. An efficient multi-feature fusion tracking method was proposed. The article presents the integration of color distributions into kernel particle filtering(KPF) framework, which has typically been used in combination with edge-based image features. The KPF invokes kernels to form a continuous estimate of the posterior density function. Kernel particle filter reasonably allocated particles by improving sampling efficiency. Experiments results show that other features can still stable and reliable track targets when a feature loses identification capabilities of target in the background clutter. Algorithm is simple and high robustness. It can be effectively applied to track target in the complex context.