抗遮挡视频图像目标实时跟踪的仿真研究  

Simulation of Anti-occlusion Arithmetic in Real-time Tracking of Video Objects

作　　者：赵林[1] 冯燕[1] 吕维[1] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710129

出　　处：《计算机仿真》2011年第5期258-262,共5页Computer Simulation

基　　金：国家自然科学基金资助项目(60772133)

摘　　要：在视频运动目标跟踪中,遮挡的出现会使所跟踪目标的尺寸和色彩等外观线索失去可靠性,容易造成误识别,进而导致对于目标的错误跟踪。为了克服这一问题,提出了一种基于目标状态预测和局部光流扫描的抗遮挡跟踪算法。算法根据卡尔曼滤波和目标颜色特征信息,预测各目标是否处于遮挡状态,在目标处于遮挡的情况下,通过由局部光流扫描得到的最佳定位信息更新目标信息。在Directshow软件下的仿真结果表明,所提出算法能够在保证实时性的前提下,在运动目标被背景遮挡或被其它目标遮挡时均能实现较准确跟踪。In the tracking of moving targets in video,occlusion can make the appearance clues of the tracked targets such as the size and the colour lose reliability,and this can cause the wrong recognition and the inaccurate tracking.To overcome the problem,this paper presents an anti-occlusion tracking arithmetic which is based on the prediction of the target state and the scaning of local optical flow.Whether the target is in occlusion is predicted by employing the Kalman filtering and the colour feature.If the object is in occlusion,the object information is updated by optimal positioning information of the local optical flow scaning.Experimental results produced by the Directshow software show that the algorithm can accurately track the moving object occluded by background or by other objects under the premise of the real-time requirement.

关 键 词：运动目标 跟踪 遮挡 卡尔曼滤波 局部光流 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]