Neural network based fingerprint classification

Abstract

In the paper the classification of fingerprints into main categories is presented. The preprocessing of a fingerprint comprises the following stages: digitization, noise reduction (filtering), segmentation and skeletonization. The classification is based on the main directions of ridges in the particular fields of the fingerprint. The Kohonen self-organizing network is used to determine the number and localization of deltas-the most characteristic structures of a fingerprint.