RESEARCH AND DEVELOPMENT
Here’s Looking At You: Iris Recognition on the Move
The movie moment might sound farfetched. But technologists say the fundamental process of using standoff iris scanners and advanced computer processing to identify individuals moving in crowds is within reach. In a few years’ time, clearing security checkpoints or customs at airports could entail walking nonstop through a designated passageway with fellow passengers, they say.
Just like fingerprints, no two irises are the same. The patterns and features of the colorful connective tissue inside the eye vary from person to person and even from eyeball to eyeball. Iris scanners typically employ near-infrared light to capture those textures, shapes and grooves. Computer algorithms developed by John Daugman, a University of Cambridge professor of computer vision and pattern recognition, convert those details into electronic codes for database comparison.
Unlike fingerprints and other biometrics, the iris remains largely immutable to physical changes caused by normal aging processes and environmental factors. Its steadfast characteristics make it one of the most accurate biometric measures, experts say.
The Defense Department has been using handheld iris recognition devices to help identify some 2 million fighting-age men in Iraq and Afghanistan. In approximately 37 countries and counting, national ID programs increasingly are including iris recognition as part of their biometrics databases, which often include fingerprints and faces.
In Tokyo, an iris recognition system identifies condominium residents and calls an elevator to take them up to their respective floors. And in 2002, iris recognition helped to identify an Afghan woman as the enigmatic orphan whose piercing sea-green eyes stared from the cover photograph of the June 1985 edition of National Geographic Magazine.
But despite the reliability, the widespread adoption of iris recognition and identification has yet to happen. Part of the problem is political in nature, with fears of personal privacy being violated. The other part of the problem is technical. Scanners traditionally have been bulky and cumbersome to use. Many controlled access systems require people to do the “iris dance,” that is, to stand close to the sensor and line up the eye with the camera so that it can capture a proper image.
“We saw that as the biggest obstacle to adoption by the market,” said Mark Clifton, vice president of products and services at SRI International Sarnoff, a Princeton, N.J.-based research firm that was one of the first to develop iris recognition.
“We have been working for a number of years on developing technology to allow the user to get his iris scanned without having to do anything except keep the eye open,” he said.
The company has produced a family of iris recognition systems that photograph the eyes as people walk through a portal or as they glance over at a checkpoint. The “iris-on-the-move” technology involves strobing near-infrared light to capture iris images quickly. The process is painless and invisible to the naked eye.
“The iris capture is no more than a picture of your iris. It’s like a picture of your face,” said Clifton. Identifying a person takes a matter of seconds. The portal technology has a throughput rate of 30 people per minute. But experts believe that can be improved.
“It’s not an unconstrained environment. You’re still putting people through a funnel with their eyes open,” said Clifton. “We are working on something like ‘The Minority Report,’ to capture anybody in a room, or walking down a mall.”
But to get to the point where sensors are actively scanning dozens of people at a time and identifying them within seconds will require more development. There are challenges inherent in imaging an iris when it is far away and not looking directly at a sensor.
“Right now, we have to capture somebody looking pretty straight on with the camera. If you get off-axis, and you get only a partial iris signature, how do you reconstruct the rest of the structure?” said Clifton. There are efforts to take partial iris images from multiple cameras and to piece them together to produce a match.
“It’s capturing enough information from the iris and doing enough processing to reconstruct the total iris — that’s really the challenge,” he said.
With standoff iris scanners maturing, there is now pressure on computer scientists to reduce the level of resolution expected by the algorithms. “Previously, I expected and got a resolution of
100 pixels in iris radius,” said Daugman. “Now, for iris images acquired from a distance measured in meters, there is pressure to accept resolutions of only about 50 pixels in the iris radius. This is challenging.”
If progress is made on both fronts, scanning and identifying irises in an unconstrained environment could become reality in the next three to four years, Clifton said. But the toughest hurdle could be overcoming fear of “Big Brother.”
Though the newer technologies feel more transparent and fluid, which is positive for user acceptance, the public still has concerns about privacy, said Daugman. “These developments are perceived as Orwellian because they allow iris cameras to be unseen, and to operate in a non-consensual surveillance mode,” he said.