Integration of Classifiers Working in Discrete and Real Valued Feature Space Applied in Two-Way Opto-Electronic Image Recognition System

K.A. Cyran (Poland)


Rough sets, computer generated hologram, evolutionary optimization, pattern recognition, hybrid systems, and neural networks


The paper presents hybrid, high speed, opto-electronic image recognition system. As in the case of classical pattern recognizers, its architecture can be decomposed into feature extractor and classifier. The feature extraction part is built as a grating based diffractive optical variable element being the computer generated binary hologram. Since such feature extractor can be produced with relatively low costs from computer generated high resolution masks, it is possible that such masks can be automatically designed specifically to given recognition task. To enable such process, the method of evolutionary optimization of the feature space has been proposed by the author in earlier works. This paper present briefly latest enhancements in this optimization method, designed predominantly for the use with neural network based classifiers. Mentioned enhancements are made possible by the original author’s modification of the notion of indiscernibility relation in rough sets theory. Together with previously described optimization method, dedicated for rough classifiers, it made possible to use as a classifier a system integrating artificial neural network and rough set based classifier. Such enhanced, not presented elsewhere system, is presented in the paper.

Important Links:

Go Back