GTC-DC 2019: Fantastically Fast Fingerprint Forensics
Note: Viewing this video may require joining the NVIDIA Developer Program or login in
GTC-DC 2019: Fantastically Fast Fingerprint Forensics
Mark Walch, Sciometrics
We’ll discuss the new technology developed by Sciometrics, in conjunction with Zebulon Tech and LatentSleuth, to address the difficulties faced by law enforcement as they process and match latent fingerprints. These are typically sparse, distorted, degraded, or disguised, and take days or weeks to match. Contactless fingerprints -- fingerprints obtained using a smartphone camera -- must be matched against contact scanning’s legacy references. To be useful, this matching process must work in real-time against a large reference corpus. LatentSleuth matches fingerprints by ridge pattern, unlike the minutiae approach used by most automated fingerprint matchers. We’ll explain how Sciometrics transitioned their CPU algorithm to NVIDIA GPUs using a Zebulon-designed GPU algorithm, and went from 10 calculations in 5 seconds to 3 billion in less than a second -- an 83,000% process improvement.