shutterstock 11395664 fmt1

MAY 2013

Motor vehicle administrators look to facial recognition technology to weed out bad actors.

BY Elizabeth Ecker

Fingerprints: They have long been the gold standard for identifying a single person as him or herself in the context of investigating and enforcing crimes, including driving offenses.

Now, add a new technology—already in place at dozens of motor vehicle administration operations—that takes what fingerprinting does and literally gives it a face: facial recognition software.

This type of biometric screening is making waves globally because of its benefits for transportation security, border control and for cross reference of mug shots for individuals with multiple arrests. Today, it’s also making major inroads for its uses among motor vehicle administrators.

Facial recognition is not without its challenges, say those using the technology. But with the proper testing and training, it has recorded more than 2,500 arrests and resulted in an additional 5,500 administrative actions in New York state alone.

“The big thing that biometrics deliver on is binding,” says Geoff Slagle, director of identity management for AAMVA. “It’s a binder between a person and [his or her] identity. How do you know the person who hands an ID to you is actually connected to it? Most of time you don’t have an answer.”

In his role at AAMVA, Slagle has worked toward increased and effective adoption of facial recognition technology among the association’s many jurisdictions. “Our community faces a daily barrage of people coming in and saying, ‘I’m here to do business, [here’s my ID].’... Imposter fraud is at an all time high. Facial recognition will go a long way to helping [combat it]. There’s no 100 percent cure, so the best we can do is improve and match more of the bad guys out there,” he says.

Finding a match

Facial recognition technology has advanced substantially since its early iterations.

Pinellas County, Fla., located west of Tampa, first began collecting digital mug shots in 1994—ahead of its peers. When technology became available to move those photos to a database where they could be compared against each other with the single click of a computer mouse, the transition was natural, says Scott McCallum, system analyst for Pinellas County Sheriff’s Office.

“We book about 60,000 people per year,” McCallum says. “Before 2001, we saw about one to two frauds in our system each year based on identity theft. Since then, we haven’t had any.”

Pinellas County uses the technology to identify incoming bookings by associating them with all other matches in the system. It also identifies people as they are leaving a jail or prison to ensure one inmate isn’t trying to impersonate another in order to be released under a different name.

The web-based system now includes other counties in Florida and amounts to about 22 million different faces. A user can input a new face into the system with an image captured either from a still picture taken by a camera or from video footage, and within seconds the program will gather any “matches” that resemble the same face.

“The outcome you are expecting is to get a ‘no match,’” says Cathy Tilton, vice president of standards and technology at Daon Solutions, a company focused on the delivery and implementation of biometric programs. “The exception case is when you do get a match.”

In those cases, there is strong potential for fraud—either a person has used a different name, or, as it applies to motor vehicle administration, may hold a license in another state.

There are also a few common reasons why a person may show up in the system more than once, prompting an “administrative match.” These include a change in name, residency or marital status. But those cases still require manpower to resolve: A person within the motor vehicle administrative department must identify the error and correct it.

Room for human error

The technology is only as good as those who operate it, as there is a strong administrative component involved. National organizations have developed a set of standards to help administrators maximize the potential of biometric screening.

“It’s a slow continuous evolution of technology in the face of some challenging problems,” says Patrick Grother, a scientist in the Information Access Division for the National Institute of Standards and Technology. “Facial recognition is upset by imaging problems.”

Those problems mostly center on a few common human errors introduced when capturing the image, including pose, illumination and expression (PIE), followed by blur and focus. Pose is the most important factor in maintaining the image standard set by NIST: A person’s face must be balanced, not tilted, with about 15 degrees of leeway. The photo also needs to be adequately lit, and the person’s expression should be as neutral as possible—not smiling or frowning.

“DMVs, to their credit, have dedicated practices and equipment that try to get good photographs,” Grother says. “This is an essential thing to do.”

Tricking the system can happen in a couple of different ways. First, problems arise when the person being photographed changes his or her expression substantially, i.e., to a heavy frown or a beaming smile. The individual can also fool the system by not looking at the camera or by altering his or her appearance by wearing a wig or heavy makeup.

However, because photos typically are taken in motor vehicle offices under the watchful eye of an administrator, these problems can be prevented easily, Grother says.