In automated and manual fingerprint minutiae matching, there exists an assumption that the location of fingerprint minutiae are somewhat random and unique to an individual's finger. However, the published estimates on the uniqueness of fingerprint minutiae configurations vary significantly. In this paper, we present an empirical analysis of fingerprint minutiae distributions as measured from a local, minutiae centric, coordinate system. Given a reference minutia, we depict the relative distribution of neighboring minutiae with 2D probability maps. Using ground truth data from NIST Special Database 27 we observe that neighboring minutiae are not randomly dispersed but prefer an inter-ridge distance of four and have non-isotropic distributions of relative minutiae orientation and relative minutiae position. Finally, based on these neighborhood probability maps we describe a new approach for estimating the uniqueness of a given minutiae configuration. Results are presented showing the relationship between the likelihood of a particular configuration and the number of minutiae in the configuration for both rolled and latent impressions.
展开▼
