Dynamic Vision, Graduate Course

Winter 2008

General Information

The use of cameras as sensors in technical systems is becoming more and more common. At the same time computer vision techniques have evolved rapidly over the past decades. The combination of a camera and a computer vision system opens up for very interesting (virtual) sensors. The measurements from these systems will be more and more commonly used in integrated systems, implying that engineers working with control, sensor fusion and related areas will need some knowledge about how these measurements are computed.

The aim of this course is to describe how we can pose and solve various estimation problems based on camera images and how cameras can be used together with other sensors.

Contents

• Rigid body motion
• Camera models
• Camera calibration
• Feature extraction
• Feature tracking
• Epipolar geometry
• Sensor fusion using cameras
• Industrial applications

Organization and Examination

The course gives 6 hp (you can receive an additional 3 hp by carrying out a project). It will be synchronized and to a certain part co-lectured with the PhD course on robotics (there are interesting possibilities for carrying out projects involving both robots and dynamic vision).

• Lectures: 5
• Bonus lecture, introduction to state estimation: 1
• Guest lecture from industry: 1
• Home work assignments: 3

The examination consists in solving 3 home work assignments.

Course Literature

• Ma, Y., Soatto, S., Kosecka, J. and Sastry, S.S. An invitation to 3-D vision - from images to geometric models, Springer, 2006.
• Recommended side reading

  Hartley, R. and Zisserman, A. Multiple view geometry in computer vision, Cambridge university press, second edition, 2003. This is a solid reference on the underlying mathematics.

  Trucco, E. and Verri, A. Introductory techniques for 3-D computer vision. Prentice Hall, 1998.

  Faugeras, O. Three-dimensional computer vision - a geometric viewpoint. MIT Press, 1993. Well-written, with good intuition. A bit older, still very readable.

  Marr, D. Vision, Freeman, 1982. A classic in computer vision.
  

Prerequisites

Basic undergraduate courses in linear algebra, signal and systems, statistics. The PhD course in sensor fusion is a plus. However, the bonus lecture on nonlinear state estimation with cover the most important concepts needed for this course.

Contact Person

Dr Thomas Schön, tel 013 - 281373, email: schon_at_isy.liu.se.