Fingerprint Classification: An Approach Based on Singularities and Analysis of Fingerprint Structure

Abstract

In this paper, we propose a new approach to fingerprint classification based on both singularities and analysis of fingerprint structure. Fingerprints are classified into five categories: arch, tented arch, left loop, right loop and whorl. The directional image is firstly divided into nonoverlapping regions to give a synthetic representation. Then, singularities (delta and core points) are extracted from a fingerprint based on the directional image with a Poincare index method. Combined with the number and location of the detected singularities, the fingerprint structural information can be exploited to classify the fingerprints. In our experiments, we focus on the fingerprint structural information to classify fingerprints. The method is invariant to rotation, translation and small amounts of scale changes. Furthermore, it does not require any iterations or feedback. Experiments have been executed on NIST4 databases.