摘要
目的目标跟踪是计算机视觉领域重点研究方向之一,在智能交通、人机交互等方面有着广泛应用。尽管目前基于相关滤波的方法由于其高效、鲁棒在该领域取得了显著进展,但特征的选择和表示一直是追踪过程中建立目标外观时的首要考虑因素。为了提高外观模型的鲁棒性,越来越多的跟踪器中引入梯度特征、颜色特征或其他组合特征代替原始灰度单一特征,但是该类方法没有结合特征本身考虑不同特征在模型中所占的比重。方法本文重点研究特征的选取以及融合方式,通过引入权重向量对特征进行融合,设计了基于加权多特征外观模型的追踪器。根据特征的计算方式,构造了一项二元一次方程,将权重向量的求解转化为确定特征的比例系数,结合特征本身的维度信息,得到方程的有限组整数解集,最后通过实验确定最终的比例系数,并将其归一化得到权重向量,进而构建一种新的加权混合特征模型对目标外观建模。结果采用OTB-100中的100个视频序列,将本文算法与其他7种主流算法,包括5种相关滤波类方法,以精确度、平均中心误差、实时性为评价指标进行了对比实验分析。在保证实时性的同时,本文算法在Basketball、Dragon Baby、Panda、Lemming等多个数据集上均表现出了更好的追踪结果。在100个视频集上的平均结果与基于多特征融合的尺度自适应跟踪器相比,精确度提高了1. 2%。结论本文基于相关滤波的追踪框架在进行目标的外观描述时引入权重向量,进而提出了加权多特征融合追踪器,使得在复杂动态场景下追踪长度更长,提高了算法的鲁棒性。
Objective Visual tracking is an important research direction in the field of computer vision and is widely appliedin intelligent transportation,human-computer interaction,and other areas.Correlation filter-based trackers( CFTs)haveachieved excellent performance due to their efficiency and robustness in tracking field.However,the design of a robusttracking algorithm for complex dynamic scenes is challenging due to the influence of lighting,fast motion,backgroundinterference,target rotation,scale change,occlusion,and other factors.In addition,the selection and presentation of fea-tures are constantly used as the primary considerations in establishing a target appearance model during tracking.Toimprove the robustness of the appearance model,many trackers introduce gradient feature,color feature,or several othercombined features rather than a single gray feature.However,they do not discuss the role of each feature and their relation-ships in the model.Method The research on correlation filter theory achieves remarkable improvements.On the basis ofthis research,the appearance model is used to represent the target and verify the observation.This process is the mostimportant part of any tracking algorithm.Moreover,the features are fundamental and difficult in appearance representation.Therefore,this study mainly focuses on the selection and combination of features.Gradient feature,color feature,and rawpixel have been discussed in previous works.As a common descriptor of shape and edge,gradient feature is invariable intranslation and light and performs well in the tracking scene of deformation,light change,and partial occlusion.However,the gradient feature of the target is not evident,and the description capability of the feature is weakened when considerablenoise is encountered in the background,target rotation,and target blur.The color of the target and background can bedistinguished although they are usually different.On this basis,a new tracking method called weighted multi-feature fusion( WMFF)tracker is proposed vi
作者
陈莹莹
房胜
李哲
Chen Yingying;Fang Sheng;Li Zhe(College of Computer Science and Engineering,Shandong University of Science and Technology,Qingdao 266590,China)
出处
《中国图象图形学报》
CSCD
北大核心
2019年第2期291-301,共11页
Journal of Image and Graphics
基金
国家自然科学基金项目(61502278
61502277)~~
关键词
相关滤波
外观描述
特征融合
加权特征
实时追踪
correlation filter
appearance description
feature fusion
weighted feature
real-time tracking
