Fingerprint Minutiae Selection for Fast Identification

In this paper the authors propose a fingerprint identification algorithm based on local minutiae structures. A high speed of identification is achieved by selecting additional features for local minutiae structures classification. Splitting by classes is determined at the stage of fingerprint processing. Therefore, the minimal amount of minutiae pairs are selected for identification and a candidate list is formed. Obviously mismatched pairs are excluded from consideration. Further minutiae pairs from the candidate list are used as seeds to develop the fingerprint fragments and for further matching. For testing the proposed algorithm about 6000 images from FVC 2000, 2002, 2004 and 2006 databases were used. The noisiest images were selected. The proposed algorithm allowed the authors to increase productivity above one million matchings per second. The algorithm was tested by FVC onGoing and showed the error value of FMR10000 less than 6.2% for a palmprint.