Date of Award

2010

Degree Type

Thesis

Degree Name

M.Sc.

Department

Computer Science

First Advisor

Boufama, Boubakeur (School of Computer Science)

Keywords

Artificial Intelligence.

Rights

CC BY-NC-ND 4.0

Abstract

This thesis addresses the relationship between camera configuration and 3D Euclidean reconstruction. Simulations have been conducted and have shown that when error is present, the larger rotation angle, the worse the reconstruction quality. When rotation is avoided, errors in the intrinsic parameters do not affect the 3D reconstruction in a significant way. Therefore, it is suggested to minimize or avoid rotation when constructing a stereo vision system. Once this configuration is applied, inaccurate intrinsic parameters, even without the prior information of intrinsic parameters, can also yield good reconstruction quality. The configuration of pure translation also provides a framework, which can be used to compute elements of intrinsic parameters with an additional geometry constraint. The perpendicular constraint is selected as an example. Focal length can be recovered from this constraint by assuming the principal point is the centre of the image.

Share

COinS