实时目标跟踪中自适应尺度计算  

Real-time object tracking based on self-adaptive scale computation

作　　者：侯小毛 王淑娟[2] 徐仁伯[3] HOU Xiao-mao WANG Shu-juan XU Ren-bo(School of Electronic Information, Hunan Institute of Information Technology, Changsha 410105, China Department of Information Technology, College of Information and Business, Zhongyuan University of Technology, Zhengzhou 450007, China School of Physics and Electronics,Central South University, Changsha 410105, China)

机构地区：[1]湖南信息学院电子信息学院,湖南长沙410105 [2]中原工学院信息商务学院信息技术系,河南郑州450007 [3]中南大学物理与电子学院,湖南长沙410105

出　　处：《计算机工程与设计》2017年第5期1314-1318,共5页Computer Engineering and Design

基　　金：湖南省教育科学规划课题基金项目(XJK015CXX005);湖南省教育厅科学研究基金项目(13C652)

摘　　要：为能够更好地解决复杂视频中目标的大尺度变化问题,提出一个在跟踪检测框架下的尺度计算算法。主要把目标分为4部分,分别计算它们的尺度因子,通过基于颜色属性的核相关滤波找出各部分最大响应位置。为解决定位不准确的问题,采用加权系数移除异常匹配点,将常规分类器所期望的训练输出转换成其它可操作的形式。仿真结果表明,该算法性能胜过其它现有算法,能够达到实时处理的要求。To solve the problem of big scale variation in complex video, a scale computation algorithm based on tracking in detec-tion was proposed The tracking object was divided into four parts,and their scale factors were computed, the maximum re-sponse location was got by adopting kernel correlation filter based on color attribution. Aiming at the difficulty of object location, weighted coefficients were used to remove the abnormal matching points and the desired training output from common classifier was transformed into other forms. Results of simulation show that the proposed algorithm outperforms other algorithms for color video with big scale variation.

关 键 词：相关滤波 实时目标跟踪 尺度计算 核变换 加权系数 颜色属性 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]