摘要
为改善Mean Shift算法的跟踪性能,分析了Mean Shift算法跟踪局限性,对Mean Shift算法流程进行了改进。Harris特征角点具有对光照、旋转、部分仿射变化以及噪声干扰具有很好的鲁棒性的特性可解决Mean Shift算法在背景过于复杂时的跟踪失败问题。Surf算法则具有对旋转、尺度缩放、亮度变化保持不变性,检测和匹配速度快等优点,能辅助Mean Shift算法在帧速过快情况下解决跟踪失败问题。提出了融合Harris角点和Surf算法的改进型Mean Shift算法。实验表明改进后的算法改善了Mean Shift算法在背景复杂以及帧速过快情况下的跟踪性能。
To improve the tracking performance of Mean Shift algorithm, the limitations of the Mean Shift algorithm are studied, the process of Mean Shift algorithm is improved. The good robustness of Harris corner in light, rotation, some affine change, and the noise could solve the tracking failure of Mean Shift algorithm in complex background. And the good characteristics of Surf algorithm in rotation, scaling and brightness as well as quick detection and matching speed could solve the tracking failure of Mean Shift algorithm in fast frame rate. A new Mean Shift algorithm is proposed. This new algorithm merged the Harris corner and Surf algorithm into the original Mean Shift algorithm. Experiments show the improved algorithm improve the tracking performance of the Mean Shift, when background is complex and the frame rate is fast.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第6期2062-2066,共5页
Computer Engineering and Design
基金
武汉工程大学第三届研究生教育创新基金项目(CX201133)
2011年湖北省教育厅青年基金项目(Q20111504)
作者简介
作者简介:杨辉(1988-)。女,湖北天门人,硕士研究生,研究方向为图像处理;E-mail:lorime@126.com
刘军(1975-),男,湖北武汉人.博士,副教授,CCF会员.研究方向为多媒体安全。