A Similarity Score for Fingerprint Images

• Activity Number: 337
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Committee of Representatives to AAAS
• Author(s): Donald Gantz*+ and Mark A. Walch and Daniel T. Gantz and John J. Miller
• Companies: George Mason University and Gannon Technologies Group and Gannon Technologies Group and George Mason University
• Keywords: Fingerprints ; Similarity Score ; Forensics

Abstract

This talk presents a similarity score for a pair of fingerprint images that is based on a novel quantification of the images. The processes presented are those used in an automated system that provides a latent fingerprint examiner with an accurate overlay of a latent (crime scene) print to each of the fingerprints that have been returned by an AFIS (Automated Fingerprint Identification System) search of a data base. The similarity score is an assessment of the accuracy of the nonlinear, invertible Warp which yields the overlay of the latent onto each returned print. The similarity score provides a prioritized ranking of all AFIS returned prints. This process is also applied independently to small snippets from a latent fingerprint and the resulting similarity scores are fused to yield an overall score for the latent fingerprint. In support of latent print examination, the system makes possible a much greater number of AFIS returned prints and provides a substantial starting point for latent to reference print examination. In order to create a Warp between two fingerprint images, a new concept of Ridge Specific Markers is introduced to serve as landmarks for constructing a Warp.