基于FOE和改进MCMC的视频运动目标跟踪方法

摘要

Modified MCMC ( Markov Chain Monte Carlo) particle filters tracking algorithm can track multi-targets steadily. But it may either miss targets or track wrong targets in dynamic scenes. Taking into account that FOE (Focus of Expansion) plays an irreplaceable role in estimating camera movement, a simple estimation model of the target positions in video with FOE is constructed first and through it the target positions are estimated. And by combining MCMC and FOE, the phenomenon of target missing and jitter are improved so that the goal of further precision in target estimation is achieved. It is shown by the experimental results that the proposed approach has good effects on dealing with target tracking problem under front or back camera translation movement.%改进的马尔可夫链蒙特卡洛MCMC( Markov Chain Monte Carlo)粒子滤波跟踪算法可以实现稳定跟踪多目标的目的.但在运动场景下,常常出现跟丢或者误跟的情况.考虑到相机聚焦中心FOE( Focus Of Expansion)在估计摄像头运动方面有不可替代的作用,首先通过构建FOE与目标在视频中位置的一个简单估计模型,估计目标的位置,再通过FOE与MCMC的结合,改善了目标丢失和抖动的现象,达到更加准确估计目标的目的.实验表明该方法对摄像头前后平移运动有比较理想的效果.