A fingerprint matching algorithm based on alignment using LPD and GCD minutia descriptors

Although various algorithms have been proposed, accurate fingerprint matching remains an unresolved problem. This paper describes a minutiae based fingerprint matching algorithm that uses minutia descriptors to help finding the optimal global transformation between two different fingerprints quickly. Two different minutia descriptors, the local patch descriptor (LPD) and the geometrical configuration descriptor (GCD), are used in this paper. A small set of matched pairs between two different fingerprints are found first based on these minutia descriptors. A rough global transformation will be found in these small set of matched pairs by RANSAC algorithm, then a proposed method of verifying the matching result is applied to filter matched pairs that extracted from RANSAC algorithm. The final global transformation will be overdetermined by the filtered matched pairs. The matching result will be calculated by the final global transformation. Experimental results show that the occurrence of misalignment is dramatically reduced and that matching accuracy is improved at the same time.