{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

lec11 - Lecture 11 HW3 due Oct 6 Questions Last Class...

Info iconThis preview shows pages 1–6. Sign up to view the full content.

View Full Document Right Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 1 Lecture 11: Sept 29, 10 HW3 due Oct 6 – Questions? Last Class Global correspondence methods Dynamic programming Graph cut methods – Two missing slides added to class notes Integration with segmentation Today’s objective Structure from Motion
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 2 Swap and Expansion Algorithms Outer loop is termed a cycle, inner loop an iteration. Finite number of cycles are needed, experiments indicate most improvement occurs after the first cycle
Background image of page 2
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 3 Applications of Stereo Navigation Mars rovers For terrestrial applications, trade-off with laser ranging systems Aerial mapping Moving aircraft naturally provides multiple images Satellite imagery: typically has a line sensor, motion is used to create the 2-D image, multiple views are acquired from different orbits of the satellite. Has been most effective for terrain mapping; built up areas are more challenging due to large disparity changes People detection in videos Industrial applications Many commercial systems available, e.g. http://www.pointgrey.com , http://www.tyzx.com
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 4 Limitations of Stereo Range accuracy falls off with distance; limitations on acceptable baseline for practical applications Accurate matching in complex scenes remains difficult Global correspondence methods tend to be slow
Background image of page 4