Hybrid framework for identifying partial latent fingerprints using minutiae points and pores

Latent fingerprints play a pivotal role in the forensics for the investigation of the crimes. Mostly, latent fingerprints found at the crime scene are partial prints. Thus, fingerprint features required for identifying the latent fingerprint is not always available. Approaches for reconstructing the partial latent fingerprints to fill the missing area is highly probabilistic. As a result, numerous false features may be extracted during feature extraction, which can affect the identification accuracy. Thus, the paper aims to propose a hybrid framework for identifying partial latent fingerprints using minutiae points (level 2) and pores (level 3), which could increase the identification accuracy. Results are evaluated on the CSRC Latent Fingerprint Touch-less Database created using Reflected Ultra Violet Imaging System (RUVIS), which shows improvement in the Rank-k identification accuracy when similarity scores of both minutiae and pores are combined. Moreover, an analysis of the identification accuracy for the number of minutiae points available in the partial latent fingerprints shows that pores could help in identifying partial latent fingerprints that have less than 5 minutiae points.