融合HOG与颜色特征的长时运动目标跟踪方法  

Long-time moving target tracking method integrating HOG and color feature

作　　者：陈强 尚振宏[1] 刘辉[1] CHEN Qiang;SHANG Zhenhong;LIU Hui(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500

出　　处：《传感器与微系统》2020年第3期76-79,90,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61462052)。

摘　　要：针对彩色的视频序列,传统的核相关滤波(KCF)跟踪方法没有考虑其中的颜色信息,且缺少遮挡处理方案,当目标在遮挡、光照变化等因素的影响下,不能很好地处理。本文结合HSV空间颜色信息,提出了一种多特征融合和对目标的丢失进行处理的跟踪方法,并且采用重检测的方法来应对目标丢失的问题。选取了公开测试视频集中的23段具有一定挑战性的视频序列,经过试验测试后,对比经典的KCF方法,平均中心位置误差(CLE)减少了29.51像素,平均距离精度提升了23.87%。实验结果表明:在目标发生快速运动、形态变化、遮挡、旋转等复杂情况下均具有较强的鲁棒性。Aiming at color video sequences,conventional kernel correlation filtering(KCF)tracking method does not take the color information into consideration and lack of occlusion processing scheme,when the target is under the influence of occlusion,illumination changes,etc.,it can not handle these very well.In order to solve this problem,combined with the HSV spatial color information,a tracking method using multi-feature fusion and the losing handling is proposed,and adopt the method of re-detection to deal with the problem of losing of the target.23 challenging video sequences in the public test video sets are selected,after the test,compared with the classical KCF method,the mean center location error(CLE)is reduced by 29.51 pixels and the average distance precision is improved by 23.87%.The experimental results show that the proposed method has strong robustness while the target is under complex conditions such as fast moving,shape changes,occlusion and rotation.

关 键 词：目标跟踪 核相关滤波 遮挡处理 特征融合 重检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]