Pattern Classification Using Correlation with Random Mask

Publisher Website
Google Scholar
Abstract
An optical apparatus for the pattern classification consisting of two steps of training and classification was made. Holographic cross-correlation of training patterns with a random mask was used to make the filters for classification. Although the error of classification was 13% in this experiment, a computer simulation of the optical method which gave a completely correct classification showed the effectiveness of this method. Effects of the characteristics of a photographic emulsion and of the displacement errors of the patterns and correlation filters were also discussed.

This publication has 10 references indexed in Scilit: