Remote Extraction of Latent Fingerprints (RELF)

Abstract

Latent fingerprints are the kind left on objects after direct contact with a person’s finger, often unwittingly at crime scenes. Most current techniques for extracting these types of fingerprint are invasive and involve contaminating the fingerprint with chemicals which often renders the fingerprint unusable for further forensic testing. We propose a novel and robust method for extracting latent fingerprints from surfaces without the addition of contaminants or chemicals to the evidence. We show our technique works on notoriously difficult to image surfaces, using off-the-shelf cameras and statistical analysis. In particular, we extract images of latent fingerprints from surfaces which are transparent, curved and specular such as glass lightbulbs and jars, which are challenging due to the curvature of the surface. Our method produces results comparable to more invasive methods and leaves the fingerprint sample unaffected for further forensic analysis. Our technique uses machine learning to identify partial fingerprints between successive images and mosaics them.