基于改进Mean-Shift与自适应Kalman滤波的视频目标跟踪  

Target tracking based on improved Mean-Shift and adaptive Kalman filter

作　　者：周尚波[1,2] 胡鹏[1] 柳玉炯[1] 

机构地区：[1]重庆大学计算机学院,重庆400044 [2]重庆市/信息产业部计算机网络与通信技术重点实验室,重庆400065

出　　处：《计算机应用》2010年第6期1573-1576,共4页journal of Computer Applications

基　　金：重庆市/信息产业部计算机网络与通信技术重点实验室资助项目(CY-CNCL-2008-02)

摘　　要：提出一种改进的Mean-Shift和自适应Kalman滤波器相结合的视频运动目标跟踪算法。对选定的跟踪目标,采用三帧差和区域增长法分割目标并得到主颜色信息。在跟踪过程中,利用自适应的Kalman滤波器估计每一帧的起始迭代位置,再利用改进的Mean-Shift算法得到跟踪位置并作为测量值反馈给自适应Kalman滤波器,并引入遮挡率因子以自适应地调节Kalman估计参数。实验结果表明,该算法能对视频中的运动目标实现检测和连续跟踪,对遮挡也有较好的鲁棒性。In this paper,a target tracking algorithm was proposed by combining the improved Mean-Shift algorithm with the adaptive Kalman filer.For a selected moving object,frame difference and region growing methods were used to segment target,and the dominant color was extracted.In the tracking process,the initial iterative position was obtained by adaptive Kalman filter in every frame,and the tracking result obtained by the improved Mean-Shift was fed back to the adaptive Kalman filter as the measurement for correction.The estimate parameters of adaptive Kalman filter were adjusted by occlusion ratio adaptively.The experimental results demonstrate that the proposed algorithm can detect and track the moving object consecutively in video and has better robustness to occlusion.

关 键 词：MEAN-SHIFT算法 KALMAN滤波 目标跟踪 遮挡 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]