Abstract:

Existing algorithms using ridge information for fingerprint matching suffer from non-linear deformation, high computational complexity and are not scalable for partial fingerprint matching. We proposed an algorithm utilizing ridge information more effectively---by choosing representative points along the ridges and using them together with minutiae, in existing minutiae matching algorithms. A ridge point selection scheme is proposed, through which only one representative ridge point (RRP) is selected per ridge. Algorithm performance is compared with that of two other minutiae-only matching algorithms, using full and partial. The approach has two merits: significant performance improvement for partial prints and usability with conventional matchers.

We further extend the use of ridge information to fingerprint individuality. We embed ridge information into the generative models by using a third distribution for ridge types. The proposed model offers a more accurate fingerprint representation. A more reliable probability of random correspondence is then computed.

Keywords. fingerprint verification, forensics, biometrics, partial fingerprint matching.