ICRA 2003 Tutorial  
   Breakthroughs in 3D Reconstruction and Motion Analysis
   
Topic
Organizers
Schedule
 Tutorial Notes
Topic

This tutorial will present the state of the art in 3D reconstruction from multiple views, including live video. Applications of  the  techniques presented include autonomous navigation, tracking, augmented reality, assisted surgery, entertainment.
The tutorial covers geometric and algorithmic aspects of estimating motion and 3D structure from image sequences for the purpose of acquiring 3D models from video, vision based control and video analysis. Applications of this technology range from special effects (scene capture, camera motion capture, virtual insertion,  image-based modeling and rendering) to photogrammetry, surveillance, autonomous navigation, medical imaging, embedded  sensor networks, and virtual reality.

Due to tremendous progress in the field of multiple view geometry, there is now a very coherent and tight story of the subject that will be presented at the tutorial. For instance it has been discovered (by some of the organizers)  that multiple-view geometry can be studied without  the need for any tensorial notation, using just simple tools and concepts from linear algebra. These recent developments as well as presentations of the basic formulation and algorithms for omni-directional cameras will be supported by additional demonstrations of the applicability of the algorithms presented.
The tutorial will be supported by a textbook that the organizers are co-authoring, which reflects such state of the art findings.

Upon completing the tutorial, participants will be familiar with the tools necessary to capture imaging data (still or video) and process it to reconstruct the 3D shape of objects in a scene to simulate 3D (rigid) motion;
estimate and segment multiple rigid body motions and to calibrate cameras.

Target Audience: Robotics researchers interested in using vision as a sensor for control tasks (navigation, manipulation, tracking, docking etc.) and 3-D model acquisition tasks. Robotics researchers interested in sensor networks and smart environments, medical, mobile and humanoid robotics, virtual and mixed reality. This tutorial is a short version of a 3-day course given at UCLA to industry and government researchers (http://vision.ucla.edu/courses/extension).


 Organizers

Jana Kosecka, George Mason University,  Yi Ma, University of Illinois at Urbana-Champaign,  Shankar Sastry, University of  California at Berkeley, Stefano Soatto, University of California at Los Angeles, Kostas Daniilidis, University of Pennsylvania, R. Vidal, University of California at Berkeley,  O. Shakernia, University of California at Berkekely

Speakers
Jana Kosecka, George Mason University,  Yi Ma, University of Illinois at Urbana-Champaign
example1
example2
example4
example4
example6

Y. Ma
J. Kosecka
Y. Ma
J. Kosecka

Schedule

   8:30  - 8:45 Tutorial Overview: Geometric approach to structure and motion recovery (Y. Ma)
   8:45  - 9:15 Rigid body motion, imaging geometry (Y. Ma) (lecture1-2.pdf)
   9:15  - 10:00 Image primitives and correspondence (J. Kosecka) (lecture3.pdf)
   10:00 - 10:15 Coffee break
   10:15 - 11:15 Two view geometry, discrete and continuous case (J. Kosecka) (lecture4.pdf)
   11:15 - 12:00 Stratified model based reconstruction from uncalibrated views (Y. Ma) (lecture5.pdf)
   12:00 - 1:00 Lunch Break
   1:00 - 1:45 Multiview reconstruction from points and lines (J. Kosecka) (lecture6.pdf)
   1:45 - 2:30 Multiview reconstruction with scene knowledge (Y. Ma) (lecture7.pdf)
   2:30 - 3:00 Coffee break
   3:00 - 3:30 Step-by-Step model building (J. Kosecka) (lecture8.pdf)
   3:30 - 4:00 Case study  UAV's landing (J. Kosecka) (lecture9.pdf)
   4:00 - 4:30 Case study  Multiple Motion Estimation (Y. Ma) (lecture10.pdf)
   4:30 - 5:00 Questions and demonstrations
 

             Tutorial Material

           
 Tutorial material will be available at this web site few weeks before the conference.
The slides will be supported by a textbook
             that the organizers are co-authoring.