An analytical framework design for latent fingerprint reconstruction, enhancement and recognition

Abstract

Latent fingerprints captured at the crime scene plays significant role as an evidence to capture the criminals. As latent fingerprints are the accidently left skin impressions, so these are found to be with broken ridge composition, overlapped patterns and spoiled minutiae information. There latent fingerprint are of no use until the reconstruction & enhancement of their ridges, quality and minutiae information. Manual reconstruction, enhancement and recognition of latent fingerprints is much expensive, dull and time overriding procedure. So, Researchers are continuously working to design an autonomous latent fingerprint recognition system but facing various challenges like availability of spoiled ridges, less information & background noise, lack of publicly availability of latent fingerprint dataset and non-availability of any specific method for fingerprint matching. In this paper, an analytical framework is presented for the reconstruction, enhancement and recognition of latent fingerprints. Initially hybrid approach of Exemplar Inpainting and Partial Differential Equation is applied for the reconstruction of spoiled ridges. Enhancement is performed to reduce the noise value and final matching is performed using binarisation approach to recognize the criminals. The proposed concept is tested for the dataset of NIST SD-27 database with the evaluation parameters of False Acceptance Rate and Genuine Acceptance Rate.