标题：Correlation Particle Filter for Visual Tracking

报告人：徐常胜

报告时间：2018年12月17日 (周一) 8:00-10:00

报告地点：数计学院2号楼311

摘要：

Visual tracking is one of the most important tasks in computer vision because of its wide range of applications such as video surveillance, motion analysis, human-computer interaction, and vehicle navigation, to name a few. The main challenge for robust visual tracking is to handle large appearance changes of target objects over time. Although significant progress has been made in recent years, it remains a difficult task to develop robust algorithms for object state estimation for tracking scenarios with challenging factors such as illumination changes, fast motions, pose variations, partial occlusions and background clutter.

In this talk, a novel Correlation Particle Filter (CPF) is proposed for robust visual tracking, which has the advantages of both particle filters and correlation filters, e.g., robustness to partial occlusion as well as scale variation, and computational efficiency. The merits of the proposed CPF tracking method are as follows. (1) The proposed CPF tracker is robust to partial and total occlusions, and can recover from lost tracks by maintaining multiple hypotheses. (2) The proposed CPF tracker can effectively overcome the scale variation problem via a particle sampling strategy as in traditional particle filter. (3) The proposed CPF tracker can efficiently maintain multiple modes in the posterior density using fewer particles than conventional particle filters do, resulting in low computational cost. (4) The proposed CPF tracker can shepherd the sampled particles toward the modes of the target state distribution using a mixture of correlation filters, resulting in robust tracking performance.

报告人简介：

徐常胜，中国科学院自动化研究所特聘研究员，中国科学院大学岗位教授，国家杰出青年基金获得者，国家重点研发计划项目“大数据多模态交互协同关键技术”首席科学家，中国科学院王宽诚率先人才计划卢嘉锡国际团队“多媒体计算国际团队”负责人。目前是国际电子电气工程师学会会士（IEEE Fellow)，国际模式识别学会会士（IAPR Fellow）和国际计算机学会杰出科学家（ACM Distinguished Scientist），担任国际计算机学会多媒体专委会中国区（ACM SIGMM China Chapter）主席和中国计算机学会多媒体专委会副主任。担任过国际刊物“IEEE Transactions on Multimedia”和“ACM Transactions on Multimedia Computing, Communication and Applications”编委，担任过国际顶级多媒体会议“2009 ACM Multimedia Conference”程序委员会主席和2010年全国多媒体会议程序委员会主席。主持过国家自然科学基金重点项目和国际合作重点项目、973课题、中科院前沿重点项目和国际合作重点项目等，在多媒体分析，计算机视觉，模式识别，图像处理等领域发表论文300多篇，其中IEEE和ACM汇刊论文80篇，国际顶级会议会议论文40余篇。谷歌学术引用超过10000次，H-Index：57。在多媒体国际顶级会议和期刊上获得最佳论文奖10余次，包括2016年国际多媒体顶级会议ACM Multimedia的唯一最佳论文奖、2016年ACM Trans. on Multimedia Computing, Communications and Applications的唯一最佳论文奖、2016年IEEE Trans. on Multimedia的最佳论文荣誉奖、2017年IEEE Multimedia的唯一最佳论文奖等。获得过2009年中国计算机学会青年科学家奖，5次获得中国科学院优秀导师奖。