하이브리드 지문인증 시스템에 관한 연구

Abstract

Fingerprint verification in biometric technologies is the most widely-used identification method due to its accuracy and convenience. In minutiae-based fingerprint verification method, which is the most general, the accuracy of identification decreases if the quality of the acquired image is low or the size of it is too small. Even though many different methods have been suggested to solve the errors, there have been limited ways to correct the errors. Meanwhile, image-based fingerprint matching uses 'direct comparing method' instead of using minutiae so that the errors coming from minutiae do not occur. Moreover, in the same size of the fingerprint image, image-based fingerprint matching uses more comparative objects than the minutiae-based fingerprint verification method. In this thesis, we analyzed the variations of performance according to the size and the shape of comparison area in order to understand the performance of fingerprint verification using minutiae. We studied fingerprint verification methods that used both minutiae-based and image-based fingerprint matching. Also, we studied a new methodology to improve the performance utilizing both minutiae-based and ridge-based fingerprint verification methods effectively. As a result, we found that the performance was improved when the center of fingerprint was included and declined when the size of comparison area is under 64×64 Pixels. As a result of experimenting with both images and minutiae, the false acceptance rate was 2.7% to 0.8% and the false rejection rate went from 0.5% to 5.5%. In addition, the Zero False Match Rate(ZeroFMR) of the fingerprint verification using both a ridge and minutiae was reduced from 29% to 9.9%, and we found that the security-performance was improved.