Shape recognition by fuzzy distance measure

Abstract

A novel shape recognition method based on fuzzy distance measure is proposed. A fuzzy set is defined on the feature vector space according to each training class, so the distance between an unknown shape and the class centroid is measured by a fuzzy distance based on the membership function. Another significant contribution of this paper is the method of constructing membership function using the statistical features of training class. Our shape recognition method is a minimum distance method in nature, but compared with the minimum distance method using common distance measure, fuzzy distance measure improves the accuracy rate of recognition greatly. The accuracy improved either compared with KNN classifier.