Omni-directional three-dimensional stereo computer vision sensor using reflective cone mirror.

Format:

Book

Thesis/Dissertation

Author/Creator:

Lin, Shih-Schön.

Contributor:

University of Pennsylvania.

Language:

English

Subjects (All):

Computer science.

0984.

Local Subjects:

0984.

Physical Description:

144 pages

Contained In:

Dissertation Abstracts International 64-04B.

System Details:

Mode of access: World Wide Web.

text file

Summary:

A good portion of computer vision research is aimed at achieving the functionality of human vision with camera-computer integration. However, the traditional cameras that model human eyes inherit the innate shortcomings of human eyes. One of the most important shortcomings is narrow field of view. Recent research has focused on overcoming these spatial "blind-spots" to achieve even greater functionality than human eyes. This thesis introduces a series of new devices and algorithms to extend the camera capabilities beyond normal human eyes. The new device has enhanced capabilities spatially (field of view), and also recovers 3D depth information all while using the simplest possible form of convex reflective mirror previously considered unusable in such tasks. The misconception that deemed the cone shape unusable is clarified and the correct theory is presented. This opens the way for achieving wide field of view using a much cheaper and simpler mirror than previously thought possible. The cone shape does introduce its unique optical properties and we are able to provide several optical designs to minimize its impact on image quality. Further, we show that we can obtain high-resolution dense 3D depth information from all directions simultaneously. It is also easier and cheaper to adapt the new sensor to acquire omni-view multi-spectral images than other curved reflective mirrors based omni-directional cameras. Simulation and experimental results conform to our theory within defined limitations.

Notes:

Source: Dissertation Abstracts International, Volume: 64-04, Section: B, page: 1799.

Adviser: Ruzena Bajcsy.

Thesis (Ph.D.)--University of Pennsylvania, 2003.

Local Notes:

School code: 0175.

ISBN:

9780496351947

Access Restriction:

Restricted for use by site license.