Communications of the ACM
Facial Scanning Is Making Gains in Surveillance
Low speed and reliability still trouble the crowd-scanning Biometric Optical Surveillance System (BOSS) tested by the U.S. Department of Homeland Security.
Credit: Getty
The U.S. government's development of a surveillance system in which computers linked to video cameras use facial recognition to identify individuals in crowds is progressing, according to newly released documents and interviews with project researchers.
The Department of Homeland Security tested the crowd-scanning Biometric Optical Surveillance System (BOSS) in the fall, but technical specialists still cite speed and reliability issues as obstacles. The BOSS project aims to address such problems by generating far more information for computers to analyze.
The system is comprised of two towers supporting robotic camera structures outfitted with infrared and distance sensors. They capture images of the same subject from different angles, and then a computer processes the images into a three-dimensional signature compiled from data such as the ratios between various points on a person's face to be compared against data about faces housed in a watch-list database.
Tests indicate that the system is still not accurate and fast enough for use by U.S. police departments, its intended users. However, Electronic Warfare Associates' Ed Tivol and University of Louisville researcher Aly Farag believe these barriers will be overcome as computer processing accelerates. Tivol says BOSS' goal is to match faces to people with 80 to 90 percent accuracy from up to 100 meters away, with the system identifying matches in less than 30 seconds.
From The New York Times
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found