25 - 29 de Noviembre de 2002

Montevideo, Uruguay

Radisson Victoria Plaza Hotel

 
CL74
 
Invariant Pattern Recognition by Zernike Moments using a Neural Network

Guillermo Cámara Cháves
Institute of Mathematics and Computer Science, University of São Paulo
gcamarac@icmc.sc.usp.br
Zhao Liang
Institute of Mathematics and Computer Science, University of São Paulo
zhao@icmc.sc.usp.br
 
Abstract

Pattern recognition has provoked a great interest in the last decades due to the use of the computers. As a consequence, numerous engineering applications have been developed. The complexity of a pattern recognition system is high because much of the information available in the real life is presented in the form of complex patterns, suffering from linear transformations even nonlinear deformations. The objective of this paper is to develop a model for invariant pattern recognition by combining Zernike moments and Fuzzy ART. The former works as invariant feature extractor while the letter acts as a robust classifier. The considered model will efficiently recognize patterns without taking in consideration the possible variations of position, rotation and scale. The experimental results shows the model can achieve high classification accuracy.

Keywords: Pattern Recognition, Zernike Moments, ART, Fuzzy ART



Texto completo        Volver

infoUYclei 2002