Tez özetleri Astronomi ve Uzay Bilimleri Anabilim Dalı

Computer-Aided Fingerprint Recognition System

Nowadays, the increase in identity fraud shows that the traditional knowledge and token-based identity determination systems are inadequate. For this reason, the use of biometric technology in identity determination became more important. It has become a necessity to implement computer-aided recognition software for biometric characteristics that have been used by people for years to recognize each other; such as face, standing, walking, voice, and that have been proven to be distinctive; such as fingerprint and hand geometry.

Fingerprint, after its distinctiveness had been determined, became one of the main biometric characteristics used for identity recognition. Being low cost and successful, the use of fingerprint recognition in commercial applications became widespread. For this reason, fingerprint recognition systems are the most widely used biometric recognition systems. Today, fingerprint recognition systems are used in determination of criminal’s identities, in personnel recognition systems of various firms, in devices like personal computers and cellular phones, and in some automobiles. Due to this wide application area, it is necessary to design suitable protocols for different recognition applications and to improve the current approaches.

The subject of fingerprint recognition is basically a pattern recognition process. Recognition is performed by using various characteristics extracted from the fingerprints. However, fingerprint recognition is a difficult and complicated process due to the fact that different fingerprints of the same finger can change considerably. These changes occur during the transfer of the fingerprint shapes into computer media using scanners; and stem from the use of different areas of the scanner surface, different angle of rotations of the fingerprint with respect to the scanner surface, nonlinear distortions due to the elasticity of the finger, and noisy fingerprint scans due to low quality scanners or the condition of the finger skin. Hence, algorithms used in fingerprint recognition systems should be flexible enough to tolerate these variations.

In this thesis, first correlation-based, ridge feature-based and minutiae-based fingerprint recognition methods are investigated. The processes that are used in the minutiae-based fingerprint recognition method are examined under separate titles and detailed using examples from the literature. Then, a minutiae-based fingerprint recognition system including these processes is implemented. Accordingly, the effects of some processes and low quality fingerprint images on system performance are analyzed using two different fingerprint databases and the results are presented.

As a result, it is shown that the recognition rate is greatly improved by enhancement and elimination of false minutiae processes. Some low quality fingerprint images are observed to have negative effects on the performances of minutiae-based fingerprint recognition systems. Yet, about 90% recognition accuracy is obtained from the developed fingerprint recognition system.