Speeding up and enhancing a large-scale fingerprint identification system on GPU

Abstract

Fingerprint identification is one of the most common biometric feature problems which is used in many applications. Although state-of-the-art algorithms are very accurate, the need for fast processing a big database of millions of fingerprints is highly demanding. In this paper, we propose to adapt the fingerprint matching algorithm based on Minutia Cylinder-Code on Graphics Processing Units to speed up the matching. Another contribution of this paper is to add a consolidation stage after matching to enhance the precision. The experimental results on a GTX-680 and K40 tesla devices with standard data-sets prove that the proposed algorithm can be comparable with the state-of-the-art approach, and is suitable for a real-time identification application.

Keywords

• Fingerprint identification
• matching
• minutiae
• Minutia Cylinder-Code
• GPU
• parallel processing