Performance evaluation of 2D to 3D fingerprint recognition

Fingerprint has been one of the most successful biometrics applied in both forensic law enforcement and security applications. Fingerprint acquisition, for many years, has been accomplished by first pressing a finger on a hard plane and then converting the image into a digital form. Recent developments in fingerprint acquisition technology have resulted in touchless live scan devices that generate 3D representation of fingerprints. To be compatible with legacy 2D fingerprint recognition techniques, the common way to develop an Automatic Fingerprint Identification System (AFIS) using 3D fingerprints is to unwrap 3D fingerprints into 2D equivalent images, then the existing algorithms for 2D fingerprint processing can be utilized. In this paper, we present some observations of 2D to 3D (unraveled 2D equivalent images from 3D model) fingerprints recognition on a small database collected using commercial scanners. False reject rate is adopted to measure the recognition performance and a commercial software VeriFinger is used for recognition. According to our experiments, the general performance of 2D to 3D fingerprints recognition is not comparable to that of 2D fingerprint recognition, specifically, the false reject rate is very high and the number of falsely extracted minutiae is large. Besides, we have analyzed possible reasons for the poor performance and pointed out some future directions regarding 3D fingerprint technology.