Biometrics

Biometrics are unique physical characteristics, such as fingerprints, that can be used for automated recognition. Since 2004, the Department of Homeland Security’s US-VISIT program, United States Visitor and Immigrant Status Indicator Technology, has collected biometrics—digital fingerprints and photographs—to protect against identity theft and fraud. Unlike names and dates of birth, which can be changed, biometrics are unique and virtually impossible to forge.

 

 

PROJECTS

“Multimodal Human ID From Video”
Charles Schmitt, Allan Porterfield, Sean Maher, David Knowles

This project aims to define the conditions under which new multimodal approaches must operate and to assess the gap between state of the art technology and operational needs. We believe that the combination of existing state-of-the-art target identification algorithms can achieve high accuracy in videos that contains significant variation in camera-subject view angle as well as significant variation in illumination conditions. Evaluation of these algorithms under conditions to be faced by production deployments is necessary to understand the gap and guide future research.

Link to Presentation
Link to Report

“Imperceptible and Real-Time 3D Scanning for Biometric Identification and Verification”
Anselmo Lastra, Henry Fuchs, Peter Stein, Greg Welch, Ali Farsaie, Will Austin, Ping Fu, Alex Chen

Identification of individuals is critical to homeland security, and 3D face recognition promises to become a reliable and robust method of identification. Many research issues remain before 3D face recognition becomes robust enough to deploy. With this research, we propose to address two of the problems: the disturbing flashing caused by structured light (the leading 3D acquisition method), and the slow speed of acquisition. Specifically, we propose to:

  • Develop practical ways to generate Imperceptible Structured Light, a UNC invention that projects patterns and their inverse so rapidly that the subject perceives only white light, and;
  • Use commodity graphics processors as computational engines to make the 3D acquisition as rapid as digital photography.

Link to Presentation
Link to Report

REPORTS AND RESEARCH BRIEFS

“Human DNA Identity Testing Policy Report”
Sara Huston Katsanis

DNA Testing is useful for establishing biological relationships and therefore is ideally suited for resolving questions of relatedness, an important factor in many migration policies. Migrant applicants may submit DNA specimens as proof of relations with family members already settled in a destination country. Recent reports of immigration fraud into the U.S. have led to efforts to improve documentation of immigration applicants and refugees, including DNA profiling. The profiled DNA may be (a) compared to claimed relatives; (b) compared to a database of crime evidence and wanted persons profiles; or (c) uploaded into a database for future reference.

Link to Research Brief

“3D Scanning for Biometric Identification and Verification”
Anselmo Lastra, Ph.D.
Published 06/16/2010

Reliable and robust identification and verification of individuals is critical to homeland security applications such as surveillance, authorization for entry to secure areas, and passport identity verification. Traditional biometrics, such as mug shots, fingerprints, and voice recognition, have been used with some success. However, they exhibit serious disadvantages for some tasks. These three biometrics, for example, are problematic for surveillance (identification); even the traditional mug shot is difficult to use in automated surveillance applications because many factors, such as lighting and frontal visibility, cannot be controlled. A relatively new biometric, 3D facial recognition, holds great promise. This brief presents the technical background of the 3D scanning technologies, briefly surveys related biometrics that may be combined with 3D recognition, provides an overview of the major technical issues, and highlights research opportunities to overcome those issues.

Link to Research Brief

Human Identification from Video: A Summary of Multimodal Approaches”
Charles Schmitt, Ph.D.
Published 06/22/2010

The U.S. Department of Homeland Security’s (DHS) Personal Identification Systems Thrust Area focuses on developing an “accurate, contactless, near real-time capability to identify known threats…through effective, interoperable multi-biometrics capabilities” (DHS Web site, 2009). A robust ability to accurately identify people from video under a variety of real-world conditions is an important capability in this context because it is potentially one of the most accurate and widely deployable technologies that requires neither contact nor consent. Maintaining high verification and identification rates when there is significant variation in illumination, view angle, degree of occlusion (eyeglasses, scarves, beards, etc.), surgical alteration, or viewing distance continues to present challenges to building truly robust person-identification systems that operate well in less controlled imaging conditions. This brief will investigate the advantages of using multimodal systems to improve accuracy and robustness of video-based biometric identification systems.

Link to Research Brief

 

Comments are closed.