Abstract

This paper deals with one aspect of the problem of object recognition, viz., the recognition of configurations. A configuration is a three-dimensional arrangement of an arbitrary number of labeled feature points. We present a computationally simple method whereby a configuration can be recognized in an infinite variety of orthographic views after being seen in only one orthographic view. The method involves simply evaluating a polynomial and requires no depth coordinates of the feature points to be given or computed. The method can, in principle, lead to recognition polynomials for configurations viewed under perspective projections, but this has not yet been done.

© 1993 Optical Society of America

Polynomial distance classifier correlation filter for pattern recognition

Mohamed Alkanhal and B. V. K. Vijaya Kumar
Appl. Opt. 42(23) 4688-4708 (2003)

Effect of tracking strategies on the velocity structure of two-dimensional image sequences

Michael P. Eckert and Gershon Buchsbaum
J. Opt. Soc. Am. A 10(7) 1582-1585 (1993)

Atmospheric turbulence and the resolution limits of large ground-based telescopes: comment

V. I. Tatarskii and V. U. Zavorotny
J. Opt. Soc. Am. A 10(11) 2410-2414 (1993)

Circular-Fourier–radial-Mellin transform descriptors for pattern recognition

Yunlong Sheng and Jacques Duvernoy
J. Opt. Soc. Am. A 3(6) 885-888 (1986)

Zernike annular polynomials for imaging systems with annular pupils

Virendra N. Mahajan
J. Opt. Soc. Am. A 1(6) 685-685 (1984)