Latent fingerprint indexing: Fusion of level 1 and level 2 features

Fingerprints have been widely used as a biometric trait for person recognition. Due to the wide acceptance and deployment of fingerprint matching systems, there is a steady increase in the size of fingerprint databases in law enforcement and national ID agencies. Thus, it is of great interest to develop methods that, for a given query fingerprint (rolled or latent), can efficiently filter out a large portion of the reference or background database based on a coarse matching (or indexing) strategy. In this work, we propose an indexing technique, primarily for latents, that combines multiple level 1 and level 2 features to filter out a large portion of the background database while maintaining the latent matching accuracy. Our approach consists of combining minutiae, singular points, orientation field and frequency information. Experimental results carried out on 258 latents in NIST SD27 against a large background database (267K rolled prints) show that the proposed approach outperforms state-of-the-art fingerprint indexing techniques reported in the literature. At a penetration rate of 20%, our approach can reach a hit rate of 90.3%, with a five-fold reduction in the latent search (indexing + matching) time, while maintaining the latent matching accuracy.