Ridge-based fingerprint matching

The diploma thesis presents an upgrade of the FingerIdent fingerprint verification system. The current version of the system uses a minutia matching procedure for comparison of two fingerprints. In order to improve the security of the system we have implemented an additional matching procedure which is based on the use of fingerprint ridges. Algorithm inputs are lists of ridge points of two fingerprints. At the beginning the algorithm searches the initial base ridge pair and matches it. Then it calculates the overlapping region of the fingerprints and recursively matches all the other pairs of ridges. In order to tolerate non-linear distortion it calculates the transformation for each matching pair separately. It repeats the procedure for the other base ridge pairs and uses the best matching score. We have carried out the integration of the algorithm with the FingerIdent system in a way that the final matching score of two fingerprints is formed as a combination of the existent matching procedure and the newly implemented matching procedure. We have tested the upgraded FingerIdent system on four test sets from the FVC 2002 competition. Each test set contains 800 fingerprint images of different qualities and sizes. We have tested different matching scenarios. Comparison of the results shows that the system which calculates the final matching score as a weighted sum of scores of the current minutia matching procedure and the newly implemented ridge matching procedure achieves better reliability.