Building a ground-truth fingerprint dataset for proficiency testing and research

Abstract

This article outlines the process used to collect fingermarks in a controlled environment, encode, and store the fingermark images. It was important that the fingermarks resemble the type of evidence that would be found at a crime scene. This meant using a variety of substrates, development techniques, and residues. The dataset also includes non-mates; the manner in which close non-matches were found using an Automated Fingerprint Identification System (AFIS) will be discussed. An efficient workflow was created which allowed the RCMP to capture 3011 fingermark images from 62 individuals. This is the foundation of the dataset that will be used for future research and proficiency testing by the RCMP. Two vital lessons learned during the process were, the importance of thoroughly cleaning objects before use and to ensure proper quality control checks were in place, including verification through a fingerprint comparison. Since the data is being used in proficiency testing, it is currently unavailable for public use.