| Fingerprint recognition is one of the most reliable and popular personal identification methods. It has been important in criminal identification applications and is getting more attention in non-criminal identification applications. In this thesis, we have refined an indexing-based fingerprint identification system using the Delaunay triangulation [1] and evaluated its performance and robustness. The system has been tested on a database of 300 fingerprints from 30 people, illustrating fast response time, good performance and robustness. Most systems store multiple fingerprints from the same finger in the database to improve the performance. We have implemented a super-fingerprint recognition approach to reduce redundancy, and more importantly, to build fingerprints containing more complete minutiae. Testing this system on the same database demonstrates dramatic reduction on memory requirements as well as faster recognition, which are useful features for large fingerprint database. |
