Abstract
Fingerprints of a personal unique characteristic have been used as an individual identification for many years. However, a perfect automatic fingerprint identification system (AFIS) does not exist. This thesis implements an AFIS with the use of fingerprint classification and minutiae pattern matching. The former stage may reduce the time of process for the latter stage. The fingerprint can be classified into five categories: arch, left loop, right loop, whorl, and others. Then minutiae, ridge endings and bifurcations, are detected for matching. Our system is tested on 6 databases of fingerprints, such as Right28, Lindex101 from PRIP Lab at NTHU and DB1, DB2, DB3, DB4 provided by FVC2000, on a PC with Athelon XP 2500 CPU and 512 MB SDRAM running Windows XP OS. Each match takes around 0.2 seconds. Among the first 3 best ranks, we achieve 98.21% (110/112), 87.87% (355/404), 95% (76/80), 93.75% (75/80), 88.75% (71/80), and 81.25 (63/80) respectively in these 6 databases. Most of non-matched images are due to unavoidable noise.