【摘 要】
：
Correlation filter has recently attracted much attention in visual tracking due to their excellent performance on both accuracy and efficiency.However,the adopted features,such as Colors,HOG and deep
【机 构】
：
School of Computer Science and Technology,Anhui University
【出 处】
：
第六届中国计算机学会大数据学术会议
论文部分内容阅读
Correlation filter has recently attracted much attention in visual tracking due to their excellent performance on both accuracy and efficiency.However,the adopted features,such as Colors,HOG and deep features,usually include noises and/or corruptions which might disturb the tracking performance.To handle this problem,we propose a novel noise-aware correlation filter method for robust visual tracking.In particular,we decompose the input feature matrix into a “clean” feature matrix and a sparse noise matrix,and then use the “clean” feature to train the correlation filter.To optimize the proposed correlation filter,we design an efficient ADMM(alternation direction of multipliers)solver.Extensive experimental results on the OTB-2013 dataset show that the proposed approach performs favorably against state-of-the-art trackers.
其他文献
离群检测是数据挖掘领域的重点问题之一,而离群检测算法中各种参数的取值严重影响着离群算法的检测结果.特别是,当存在具有任意形状的簇和变化的密度的数据集时,在没有先验知识的情况下难以确定适当的参数.为了解决这个问题,论文在自然邻居方法的基础上,提出一种利用加权自然邻居邻域图进行离群检测的算法.该算法在整个过程不需要人为设置参数,并且在不同分布特征的数据中准确找到数据集中的全局离群点和局部离群点.人工数
Secure deduplication techniques have been wildly used in cloud storage to save both disk space and network bandwidth.However,traditional schemes store only one copy in cloud which raises problems with
In order to solve a series of problems caused by the disorderly cy-cling of bicycle-sharing,the modeling problem,forecasting the needs of the shared bicycle and selecting the location for the bicycle-
Friend recommendation is a fundamental service for both so-cial networks and practical applications.The majority of existing friend-recommendation methods utilize user profiles,social relationships,or
This paper presents a simple and efiective method to verify similar vehicle images.In order to provide a meaningful interpretation of the verification,we propose to detect the local difierences betwee
Despite recent breakthroughs in object detection with static images,extending state-of-the-art object detectors from image to video is challenging.The detection accuracy sufiers from degenerated objec
动态社区发现是研究网络演化的重要技术.然而,随着网络规模的日益增长,传统的单机算法难以处理大规模网络.此外,现有的一些社区质量衡量指标存在一定的局限性,如持久力(permanence),加权聚集系数WCC(Weighted Clustering Coefficient)等具有较高的计算复杂性,基于这些指标的社区发现算法具有较高的时间复杂度,难以应用于大规模动态网络.该文针对WCC的高时间复杂度,提
互联网的普及和网络连接设备与访问方式的多样化,为人们生活带来巨大便利的同时也带来巨大的安全挑战.网络入侵方式与手段日趋多样化且变异速度快,传统入侵检测方法在有效性、自适应性和实时性方面难以应对日益复杂网络环境的安全监控要求.本文提出一种基于在线自适应极限学习机(Online adaption extreme learning machine,OAELM)选择性学习的网络入侵检测方法(SEoOAEL
Videos spread over the Internet contain a huge knowledge of human society.Diversified knowledge is demonstrated as the storyline of the video un-folds.Therefore,realization of automatically constructi
Graph simulation as a well studied model of graph pattern matching problem,has been adopted to reduce the complexity and meet the need of novel applications such as mining potential associations betwe