A New Fingerprint Matching Approach Using Level 2 and Level 3 Features

Abstract

Fingerprint recognition is considered as one of the most popular and reliable techniques for automatic identification due to the well-known distinctiveness and persistence of fingerprints. In an automated fingerprint recognition system (AFRS), fingerprint matching is an important and essential step. The distortion and noise of given fingerprints and the inaccurate feature extraction make fingerprint matching a very difficult problem. With the advent of high-resolution fingerprint imaging techniques and the increasing demand for high security, sweat pores have been recently attracting increasing attention in automatic fingerprint recognition. Therefore, this paper takes fingerprint pore matching as an example to show the robustness of our proposed matching method to the errors due to fingerprint representation. This method using Onion Peeling algorithm of computational geometry, utilizes Level 3 features in conjunction with Level 2 features for matching. The experimental results on two databases of high-resolution fingerprints demonstrate that the proposed method can achieve much higher recognition accuracy compared with other state-of-the-art pore matching methods. The algorithm has been tested on two databases (NIST 9 database and FVC2006). We have measured the performance of our proposed algorithm in terms of receiver operating characteristics (ROC). For the NIST9 database, at ~11% false acceptance rate (FAR) the genuine acceptance rate (GAR) is ~ 92% and at ~0% FAR, observed GAR is ~74%. For the FVC2006 database at ~0% FAR the GAR observed is 71% and at ~19% FAR, GAR is 89%. It is concluded that with proposed technique we can have better GAR at low FAR with reduced computational complexity.