一网通办| 邮箱登录| English| VPN链接| 电话查询|

通知公告

澳门一沙金网址娱乐网站

信息工程学院邀请胡钊政博士学术报告的通知

 

题  目:Depth Recovery and Stratified Reconstruction from Active Vision

(基于主动视觉的深度恢复与分层重建)

主   讲:Zhaozheng Hu(胡钊政 博士), PhD

  间:2011年12月6日9:00

  点:西教二——407

Abstract: 3D reconstruction is one of the two fundamental tasks in the field of computer vision. Recently, 3D data and 3D features play more and more important role for scene understanding and motion analysis. For example, 3D data and features are emerging in various applications, such as 3D pavement cracking measurement, 3D roadway scene reconstruction to enhance road asset data collection and inventory, 3D vehicle and pedestrian detection for surveillance, 3D face recognition, etc. Existing 3D reconstruction methods usually suffer from the problems of complicated camera and system calibration, robustness, and accuracy, etc. However, 3D reconstruction can be enhanced by exploiting the prior knowledge of camera or object motions so that we can design some specific motions to serve this purpose. In this presentation, we will show how the active vision can help us to compute scene depth, reconstruct the affine and Euclidean 3D structures. The proposed method can successfully avoid the step of camera calibration for absolute or relative depth computation. The affine reconstruction doesn’t need any references points to expand the affine frame and is easily upgraded into Euclidean reconstruction. Furthermore, we investigate how the depth and the infinite homography can be linearly computed from each other so that the method can be extended to compute depth information from two general un-calibrated views. Computation of the height to a plane can also be feasible based on the proposed affine reconstruction. Error analysis model is presented to quantitatively evaluate depth computation uncertainty.

Short Biography

胡钊政,博士,副教授,日本京都大学JSPS研究员。分别于2002年与2007年在西安交通大学信息与通信工程系获学士与博士学位。2004年11月至2005年6月在香港中文大学计算机视觉实验室进行访问研究。2007年7月至2009年8月,在美国佐治亚理工学院 (Georgia Institute of Technology)从事博士后研究工作。2009年9月开始在大连海事大学信息学院担任副教授。从2010年7月开始,以日本学术振兴会(JSPS)外国人特别研究员身份前往日本京都大学从事研究工作。胡钊政博士主要研究方向集中在视觉几何,三维重建,车载系统,智能交通系统,以及多摄像机主动视觉监控系统等方面。目前已经主持或参与多项高水平科研项目,包括日本学术振兴会Grant-In-Aided(GIA)项目,美国科学院公路协作研究计划项目(NCHRP),美国佐治亚州交通部项目,辽宁省教育厅科研项目,科技部支撑计划项目,香港特区政府创新研究计划项目(ITF)等。目前已经发表论文近20篇,其中被SCI检索8篇,被EI检索14篇。胡钊政博士曾八次担任著名国际学术会议程序委员会委员(TPC),目前还担任十余家国内外著名学术期刊审稿人。

 

XML 地图 | Sitemap 地图