基于局部稀疏表示的多帧图像特征目标跟踪方法  

Multi-frame image feature target tracking method based on local sparse representation

作　　者：张博 Zhang Bo(Collegeof Information Science and Engineering,Changsha Normal University,Changsha 410100,China)

机构地区：[1]长沙师范学院信息科学与工程学院,长沙410100

出　　处：《国外电子测量技术》2022年第6期25-30,共6页Foreign Electronic Measurement Technology

基　　金：教育部中国高校产学研创新基金(2020ITA05028);教育部产学合作协同育人项目(201901014024);湖南省普通高等学校教学改革研究项目(HNJG-2021-1195);湖南省社会科学成果评审委员会一般项目(sskl202219)资助。

摘　　要：多帧图像中存在的噪声相较于单帧图像中的噪声更多,因此,为了准确跟踪多帧图像目标,提出了基于局部稀疏表示的多帧图像特征目标跟踪方法。采用局部稀疏表示方法将多帧图像目标进行分块,跟踪子块的移动轨迹与当前状态,基于当前状态构建目标函数,将经目标函数处理后的分块按照一定的顺序进行排列得到初始检测结果,在此基础上,对图像仿射变换,实现多帧图像特征目标跟踪。实验结果表明,所提方法目标跟踪时的中心误差较低,不超过5%。在10 min目标遮挡和模糊视频中,可以在2 s之内完成目标跟踪。在定性评价上,所提方法能够准确跟踪目标,有效提高图像特征目标跟踪效果。There is more noise in multi frame images than in single frame images.Therefore,inorder to accurately track multi frame image targets,a multi frame image feature target tracking method based on local sparse representation is proposed.In the mothod,the multi frame image target is divided into blocks,which tracks movement trajectory of the sub-blocks and current state,constructs the objective function based on current state and arranges the blocks processed bytheobjective function in a certain order to obtain initial detection result,and thenon this basis,the image affine transformation realizes multi-frame image feature target tracking.Theexperimental resultsshowthatthecentererrorof the proposed method is low,less than 5%.In the 10 minute target occlusion and blurvideo,the target tracking can be completed within2s.In the qualitative evaluation,the proposed method can accurately track the target,which effectively improve trackingeffect of theimage feature target.

关 键 词：局部稀疏表示 多帧图像 分块特征 目标跟踪 仿射变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]