基于改进Mean Shift算法的实时视频目标跟踪  

Real-time Video Target Tracking Based on Improved Mean Shift Arithmetic

作　　者：胡波[1,2] 

机构地区：[1]浙江东方职业技术学院工程技术系,浙江温州325011 [2]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《光子学报》2012年第4期461-465,共5页Acta Photonica Sinica

基　　金：国家自然科学基金(No.60971074)资助

摘　　要：设计了一套嵌入式平台上实现的视频目标跟踪系统.该系统采用CMOS图像传感器获取视频信号,利用Z228多媒体芯片自带的ARM9处理器完成视频信号的控制,并通过MPEG-4硬件编码器实现视频信号的压缩.用Mean Shift算法跟踪运动目标,针对其收敛的局限性设置多个搜索点来提高其跟踪效果.通过减少采样点和标记已计算点来提高代码运行速度,增强了跟踪的实时性.实验结果表明,本系统能以27fps速率连续稳定地实现视频目标的跟踪.A video target tracking system is designed which is achieved in the embedded platform.The system acquires video signal by using CMOS image sensor,controls video signal in ARM9 processor which is obtained in Z228 multimedia chip,and compresses video signal through MPEG-4 hardware encoder.Meanwhile,the Mean Shift arithmetic is used to track moving target.Considering the convergence limitation,multiple search point is set to improve the tracking results.The operating speed of code is raised through decreasing sampling point and labeling computed points,thus the tracking instantaneity is enhanced.The experiment results indicate that this system can achieve video target tracking continually and steadily at the speed is 27 fps.

关 键 词：嵌入式 视频目标跟踪 图像传感器 Mean SHIFT算法 实时 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]