Implementation of an automatic fingerprint identification system

Fingerprints used for identification and verification has attracted many biometric researchers. Although no identical fingerprint from distinct identities was found, a perfect automatic fingerprint identification system (AFIS) may not exist. This paper implements an AFIS with the use of fingerprint classification and a minutiae pattern matching. Our system is tested on a database composed of 112300 times 300 right index fingerprint images contributed by 28 persons, each has 4 duplicates via Veridicom FPS110 reader. Each match takes 0.2 seconds on a PC with Athelon XP 2500 CPU and 512 MB SDRAM running Windows XP. We achieve 110 correct matches out of 112 tests. The ony two mismatched images are due to unavoidable noise. The result of testing our system on 4 times 101 left index fingerprints of more noisy images from 101 persons achieves 88% recognition rate.