摘要
针对于复杂场景下,跟踪的目标容易产生漂移甚至跟踪失败的情况,本文提出了一种基于AdaBoost置信图的红外与可见光目标跟踪算法。首先,以颜色和纹理特征为描述子对红外与可见光图像的目标样本与背景样本进行表征和AdaBoost分类,并基于分类度计算得到红外与可见光图像的置信图;然后,在置信图中分别计算它们的目标候选者与其模板置信图之间的相似度,并将两相似度进行加权融合,构建联合目标函数;最后,对目标函数进行泰勒展开和求导等操作,推导出联合位移公式,并运用均值漂移算法完成目标搜索。对多组红外与可见光图像序列对测试结果表明,本文提出的算法在处理光照变化、目标交汇、目标遮挡等方面都表现良好。
To address the problem that the tracker is easy to drift away from the target and even failure in complex scenes,this paper presents an infrared-visible target tracking algorithm based on AdaBoost confidence map.Firstly,the target samples and background samples in infrared-visible images are characterized by using color and texture descriptor and are classified using AdaBoost classifier,and then the confidence maps of infrared and visible images are calculated based on the classification scores.Secondly,the similarity between confidence maps of target candidate and its template is calculated for visible and infrared images,and the visible similarity and infrared similarity are integrated into a joint objective function by weighting.Finally,a joint target location-shift formula is induced by performing multi-variable Taylor series expansion and maximization on the objective function,and the optimal target location is gained recursively by applying the mean shift procedure.The experimental result in infrared-visible image sequences demonstrates that the proposed method performs well in dealing with illumination change,target intersection,target occlusion and so on.
作者
张灿龙
苏建才
李志欣
王智文
ZHANG Canlong;SU Jiancai;LI Zhixin;WANG Zhiwen(Guangxi Key Lab of Multi-source Information Mining&Security,Guangxi Normal University,Guilin Guangxi 541004,China;Guangxi Collaborative Innovation Center of Multi-source Information Integration and Intelligent Processing,Guilin Guangxi 541004,China;College of Computer Science and Communication Engineering,Guangxi University of Science and Technology,Liuzhou Guangxi 545006,China)
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2018年第4期42-50,共9页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金(61866004
61663004
61462008
61751213)
广西自然科学基金(2017GXNSFAA198365
2016GXNSFAA380146)
柳州市科学研究与技术开发工程项目(2016C050205)
