改进ViBe算法与Meanshift结合的目标跟踪方法  

An Moving Target Tracking Method Based on Improved ViBe Algorithm and Meanshift

作　　者：王凯[1] 高勇 吴敏[1] 张翔 WANG Kai;GAO Yong;WU Min;ZHANG Xiang(Research Center,The 2nd Research Institute,Civil Aviation Administration of China,Chengdu 610041,China;School of Electronic Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China)

机构地区：[1]中国民用航空总局第二研究所科研中心,成都610041 [2]电子科技大学电子工程学院,成都611731

出　　处：《科学技术与工程》2020年第21期8663-8670,共8页Science Technology and Engineering

基　　金：国家自然科学基金(U1733111,U1833101);国家重点研发计划(2018YFC0809500);四川省省院省校科技合作研发项目(2019JZ0030);四川省科技项目(20ZDYF3376,20ZHSF0359);民航空管科技项目(2018XZ-214)。

摘　　要：为解决视频监控系统中因光照变化、相似颜色干扰及快速动目标而导致的目标易丢失问题,提出一种改进ViBe算法与Meanshift结合的目标跟踪方法。首先采用三帧差分和ViBe算法结合进行前景目标提取和检测,以消除鬼影的干扰,利用色度特征和梯度特征相结合的方法来抑制阴影;同时通过将边缘特征融入到Meanshift算法中,引入运动矢量在当前帧中预测下一帧运动目标的位置,实现场面监视环境中运动目标的持续、准确跟踪。通过在监控视频中行人、车辆及飞行器等不同场面目标做实验,验证了本文方法不仅能够持续、准确地跟踪运动目标,并且可适用于场面雾天低能见度条件下的运动目标跟踪。The change of illumination,similar color interference and fast moving target in video surveillance system may cause the moving target tracking loss problem.To solve the problem,a moving target tracking method based on improved Vibe algorithm and Meanshift was proposed.Firstly,three-frame differences and ViBe algorithm were used to obtain foreground target to eliminate the interference of ghosts.Moreover,chroma and gradient features were used to restrain shadow.Finally,the continuous and accurate tracking of moving targets were realized in complex video surveillance scenes by integrating edge features into Meanshift algorithm and introducing motion vector to predict the next frame position of the current frame of moving target.Experiments on pedestrians,vehicles and aircraft in surveillance video show that the proposed method can not only track moving targets continuously and accurately,but also be applied in the foggy low visibility conditions for tracking moving targets.

关 键 词：视频监控 Vibe算法 MEANSHIFT 目标跟踪 前景目标 运动矢量 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]