The two-dimensional view, obtained with a camera, of a three-dimensional (3-D) object varies with the 3-D orientation of this object, complicating the recognition task. In this work we address the problem of estimating the pose of a 3-D object knowing only a 2-D projection. The proposed technique is based on a combination of synthetic-discriminant-function filters and neural networks. We succeed in estimating two orientations: in-plane and out-of-plane rotations within a 8 degree square range.

Date Published: 27 November 2002
PDF: 8 pages
Proc. SPIE 4789, Algorithms and Systems for Optical Information Processing VI, (27 November 2002); doi: 10.1117/12.469139

Published in SPIE Proceedings Vol. 4789:
Algorithms and Systems for Optical Information Processing VI
Bahram Javidi; Demetri Psaltis, Editor(s)