Fingerprint recognition with geometric relation of minutiae points

In this paper, two descriptors for automatic fingerprint recognition that utilize geometric relations among minutiae points are proposed. Both of the descriptors, which can also be called as geometric signatures, utilize orientation, position and type information of minutiae points. The first descriptor represents the local geometric distribution of points by using a histogram, while the second descriptor represents the same information by using a binary representation (present/absent). The experiments are conducted on the FVC 2004 [9] dataset, by using four fingerprints from each of the ten people that take part. In order to objectively assess the performance of the descriptors, minutiae points on the fingerprint images are marked manually by human operators. Additionally, in order to analyze the robustness of the proposed descriptors, the dataset is augmented by extra points that are obtained by adding various levels of noise to the original points in orientation, position and scale domains.