基于Markov跳变非线性系统的粒子滤波跟踪算法

A Particle Filter Algorithm Based on Markov Transition Nonlinear Systems

针对粒子滤波算法在红外多目标跟踪,尤其是目标被遮挡时跟踪效果不佳的问题,本文在粒子滤波算法中引入Markov跳变非线性系统,在贝叶斯估计时融入Markov随机场来描述目标交汇时的交汇模型,并将其应用在有目标遮挡情况的红外图像的多运动目标跟踪中。实验结果表明,引入Markov跳变的粒子滤波算法,其抗遮挡能力大大提高,跟踪效果优于经典的粒子滤波算法。

Considering the deficiency of particle filter target tracking algorithm in infrared,especially in problems with variable number,multi-target tracking and target occlusion,this passage introduces Markov Random Field（MRF） to describe the interaction model of multi-objective which is based on Markov transition Bayesian state estimation of nonlinear systems starting. This model accomplishes the multi target tracking under specific system and can be applied to condition of multi-target tracking with target occlusion. The experimental result shows that Markov transition particle filter algorithm exhibits a better target occlusion ability than that of traditional particle filter algorithm.