Texas Firm Thinks Facial Biometrics Will Work on Air Travelers
Trying to reduce the error rate of a one-to-many technique so that it approaches a useable level has already proven to be "very difficult."
The divergent airport-security approaches of two facial-recognition firms, Advanced Digital Imaging Research (ADIR) of League City, Texas, and Sunnyvale, Calif.-based Applications for Vision (A4Vision), likely will be a source of debate in the biometrics industry for some time to come.
At issue is whether it's better to push on with the "positive authentication" approach, where a facial image is matched to other identifying information such as an employee number, or pursue a less-well-developed approach often known as "one-to-many," where a facial image is compared to an image database for an attempted match.
The latter also is the technique that ADIR is trying to perfect in a two- week-old experiment at Waco Regional Airport (ACT) in Texas. ADIR has already assembled a "known database" of faces from its own employee files and that of its parent company, Chatsworth, Calif.-based Iris International, Inc. [IRIS]. ADIR also wants to assemble an anonymous sampling of 10,000 new faces from air travelers passing through the Waco airport. Then, by summer's end, the firm hopes to discover several new ways of refining its technology, ADIR President Ken Castleman says.
But A4Vision's CEO, Grant Evans, thinks ADIR has its work cut out for it. Trying to reduce the error rate of a one-to-many technique so that it approaches a useable level has already proven to be "very difficult," Evans tells Air Safety Week. "I applaud the company on pushing the envelope," but doubts that its efforts will be successful. Where the technique already has been tried, it has "failed dramatically." And if the latest experiment fails, the attendant publicity may hurt the rest of the industry.
When another form of biometric identifiers -- fingerprints -- are compared against a known database (like those with criminal histories), there's always a few "false positives" where sets of prints appear to match a file in the database, Evans explains. As the number of new prints being checked increases, the number of false positives also grows.
Evans says that a governmental study conducted at a large airport, and scheduled for public release within a month or so, will show that A4Vision's brand of 3D facial biometrics in positive authentication is highly accurate. Compared to error rates averaging 30 percent for several other firms' biometrics, including facial imaging, iris scanning and fingerprinting, A4Vison's error was well under 1 percent.
For his part, ADIR's Castleman also admits that prior testing of the one- to-many technique has not been successful. Waco, however, will be the first time such a large sample of new faces will be used, and the current months-long trial is only for improving the prototype, he tells Air Safety Week. In the next testing phase, possibly beginning late this year or early in 2006, ADIR hopes to check its system at airport checkpoints. In informal discussions, TSA has been supportive of ADIR's efforts so far, and Castleman believes it likely that the agency will support a "real world" trial on arriving passengers.
From signs posted throughout Waco airport explaining the project's scope and purpose, passengers can agree to be voluntary project participants. They will leave their facial images at an ADIR portal, which resembles a magnatometer, and features multiple cameras to get a composite image of subjects' faces. Although there will be a number attached to their images, ADIR will keep no other identifying information, nor will the images be passed on to any law enforcement agency. Waco Regional was chosen as the test site not only because of its proximity to ADIR's offices -- no more than 250 miles -- but also because it is about the right size for this type of testing.
- « Previous Page
- 1
- 2
- Next Page »










