基于微分光流和粒子滤波的视频运动目标跟踪  

Moving Object Tracking in Video Based on Derivative Optical Flow and Particle Filtering

作　　者：赵林[1] 冯燕[1] 贾应彪[1] 

机构地区：[1]西北工业大学电子信息学院,西安710129

出　　处：《计算机工程》2011年第13期235-237,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60772133)

摘　　要：针对传统跟踪方法中存在的背景遮挡问题,提出一种基于微分光流的粒子滤波跟踪方法。该方法建立一个用于描述目标的微分光流模型,该模型在每一帧的跟踪处理之前获得更新。使用光流相似函数计算目标与粒子所表示区域的相似度,通过粒子滤波估计实现目标跟踪。实验结果表明,该方法在目标被与目标灰度值相差较大和相近的2种背景遮挡时均能保持对目标的正确跟踪。Aiming at the problem of background occlusion existed in traditional video tracking method,a novel particle filtering tracking method based on derivative optical flow is proposed.In this method,a derivative optical flow model used for describing the object is built,and the model is updated ahead of tracking in every frame.The similarity between the object and the particle is measured by the optical flow similarity function,then the tracking is accomplished by the estimate of particle filtering.Experimental results show that the method can correctly track the object occluded by background,no matter whether the gray intensity of the object and background is similar or different

关 键 词：运动目标 跟踪 遮挡 微分光流 粒子滤波 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]