Lehigh University - Zhu seminar

Dr. Zhigang Zhu

"Content-Based Dynamic 3D Mosaics for Urban Scene Modeling and
Target Detection "

Tuesday, October 23, 4:00 PM

Packard Lab---Room 466

(Refreshments will be served starting at 3:30p.m. in the lobby of
Packard Lab)

We propose a content-based 3D mosaic (CB3M) representation for an urban
3D and dynamic scene recorded by a camera on a ground or aerial mobile
platform. The content-based 3D mosaic (CB3M) representation is a highly
compressed visual representation for a very long video sequence of a
dynamic 3D scene. For a real image sequence of a campus scene, with
1000 frames of 640*480 color images, a compression ratio of a few
thousands to more than ten thousand can be achieved, depending on what
levels of details are preserved. More importantly, the CB3M
representation has object contents of both 3D and motion information, allowing further automated target recognition, indexing and retrieval. In this talk, I will describe the two important steps in generating the CB3M
representations: 3D registration/mosaicing under motion parallax, and segmentation-based stereo matching of multiple pushbroom mosaics. Experimental results and video demos will be shown for the CB3M
representation construction of both simulated and real-world scenes (including an NYC scene).

This is a joint work with Mr. Hao Tang, Mr. Edgardo Molina and
Professor George Wolberg at the City University of New York. The work
is supported by the Air Force Research Lab under the AFRL/SN RASER
Program for Multi-Sensor Registration (Award No. FA8650-05-1-1853), and
an NSF CRI grant (Award No. CNS-0551598).

BIOSKETCH

Dr. Zhigang Zhu is currently a Professor in the Computer Science
Department, the City College of New York. He directs the City College
Visual Computing Laboratory (CcvcL), and co-directs the Center for
Perceptual Robotics, Intelligent Sensors and Machines (PRISM) in Grove
School of Engineering at CCNY. Previously he has been an Associate
Professor at Tsinghua University, and a Senior Research Fellow at the
University of Massachusetts, Amherst. His research interests include 3D
computer vision, multimodal sensing, Human-Computer Interaction (HCI),
video representation, and various applications in education,
environment, robotics, surveillance and transportation. Dr Zhu is a
senior member of the IEEE, a senior member of the ACM and an Associate
Editor of the Machine Vision and Applications Journal. He is Co-General
Chair (with Prof. Thomas S. Huang) of the 2007 IEEE Workshop on
Multimodal Sentient Computing, in conjunction with CVPR 2007 in
Minneapolis, Minnesota, and Co-Guest Editor (with Prof. Takeo Kanade)
of the up-coming Special Issue on Modeling and Representations of Large
Scale 3D Scenes, International Journal of Computer Vision.

		Dr. Zhigang Zhu "Content-Based Dynamic 3D Mosaics for Urban Scene Modeling and Target Detection " Tuesday, October 23, 4:00 PM Packard Lab---Room 466 (Refreshments will be served starting at 3:30p.m. in the lobby of Packard Lab) We propose a content-based 3D mosaic (CB3M) representation for an urban 3D and dynamic scene recorded by a camera on a ground or aerial mobile platform. The content-based 3D mosaic (CB3M) representation is a highly compressed visual representation for a very long video sequence of a dynamic 3D scene. For a real image sequence of a campus scene, with 1000 frames of 640*480 color images, a compression ratio of a few thousands to more than ten thousand can be achieved, depending on what levels of details are preserved. More importantly, the CB3M representation has object contents of both 3D and motion information, allowing further automated target recognition, indexing and retrieval. In this talk, I will describe the two important steps in generating the CB3M representations: 3D registration/mosaicing under motion parallax, and segmentation-based stereo matching of multiple pushbroom mosaics. Experimental results and video demos will be shown for the CB3M representation construction of both simulated and real-world scenes (including an NYC scene). This is a joint work with Mr. Hao Tang, Mr. Edgardo Molina and Professor George Wolberg at the City University of New York. The work is supported by the Air Force Research Lab under the AFRL/SN RASER Program for Multi-Sensor Registration (Award No. FA8650-05-1-1853), and an NSF CRI grant (Award No. CNS-0551598). BIOSKETCH Dr. Zhigang Zhu is currently a Professor in the Computer Science Department, the City College of New York. He directs the City College Visual Computing Laboratory (CcvcL), and co-directs the Center for Perceptual Robotics, Intelligent Sensors and Machines (PRISM) in Grove School of Engineering at CCNY. Previously he has been an Associate Professor at Tsinghua University, and a Senior Research Fellow at the University of Massachusetts, Amherst. His research interests include 3D computer vision, multimodal sensing, Human-Computer Interaction (HCI), video representation, and various applications in education, environment, robotics, surveillance and transportation. Dr Zhu is a senior member of the IEEE, a senior member of the ACM and an Associate Editor of the Machine Vision and Applications Journal. He is Co-General Chair (with Prof. Thomas S. Huang) of the 2007 IEEE Workshop on Multimodal Sentient Computing, in conjunction with CVPR 2007 in Minneapolis, Minnesota, and Co-Guest Editor (with Prof. Takeo Kanade) of the up-coming Special Issue on Modeling and Representations of Large Scale 3D Scenes, International Journal of Computer Vision.