基于视觉显著性的Wang-Landau蒙特卡罗采样突变目标跟踪算法

摘要

突变运动目标的鲁捧跟踪是计算机视觉领域的一个具有挑战性的问题.提出了一种基于视觉显著性的Wang-Landau蒙特卡罗采样(WLMC)跟踪算法,用于解决复杂场景下目标发生运动突变的跟踪问题.该算法首先对全局场景进行分块获取子区域,然后使用WLMC方法进行目标状态采样来跟踪发生运动突变的目标.算法将视觉显著性作为先验引入跟踪框架,提出了结合显著性先验的接受函数,通过每个子区域的显著性计算来引导马尔可夫链的构造.和以往方法相比,该算法既保留了WLMC采样方法对全局状态空间的广度覆盖性,又以视觉显著性特性引导采样,避免了全局采样的盲目性,从而提高采样效率.实验结果表明,该算法对发生运动突变的目标进行跟踪,具有良好的鲁棒性.%Robust tracking of abrupt motion is a challenge problem in computer vision.In this paper,we proposed a saliency-based Wang Landau Monte Carlo (WLMC) tracking method for abrupt motion problem in real world scenarios.Firstly,the spatial space is divided into disjoint sub regions.Secondly,a sub-region is selected randomly by the WLMC sampling method.Visual saliency as prior is introduced into tracking framework where saliency of each sub-region is integrated into Markov Chain Monte Carlo acceptance mechanism to guide effective states sampling.Therefore the method can avoid overall sampling and improve sampling effectiveness.Experimental results demonstrate that our approach samples the states of target efficiently in whole state space and outperforms several state-of-the-arts algorithm.