Faces in the crowd

 

November 18, 2015

“Have you seen this girl?” You can’t avoid her.

You’re waiting at the bus stop, looking at the billboard across the street, and she’s there. You see a tabloid on display at the grocery store, and she’s there. You turn on the television after dinner, and she’s there.

What you see is a pair of wide blue-green eyes staring out from the missing-child notice circulating like wildfire across every major news outlet. If you look closely, you’ll see a dark mark in the girl’s left iris—a key marker of her identity.

The Internet buzzes with pictures of this blond-haired toddler. Some have the catch phrase “Find Maddie”; others ask, “Who Took Our Maddie?”

It’s 2007, and the parents of Madeleine McCann have launched a global campaign in the midst of their worst nightmare: the disappearance of their three-year old daughter.

Just hours before she was gone, Madeleine played in the warm sands of Praia da Luz, Portugal. Her biggest worry? Probably that the turquoise tide would rush in and sweep away her sandcastle. But later that evening it was reported that she was abducted from her quiet bedroom where she slept wrapped in a pink blanket, next to her younger brothers.

Flash forward to today. Despite having become a household name in the immediate aftermath of her disappearance, she is still missing. Although the memories of Maddie’s face have faded from public memory, Marios Savvides hasn’t stopped thinking about her case. The Carnegie Mellon University research professor knows that technology might have uncovered what happened to her in the first few hours following her disappearance.

What he has been working on since then is an advanced facial and iris recognition system. Originally, this research was developed as a commission by the U.S. Department of Defense for military purposes. It’s now being used for broader purposes because of the adaptability of technology that is able to register the identity of any person within a 12-meter radius using their iris. It’s called the CMU Long Range Iris (LRI) system.

The system has been impressive enough to catch the attention of the Carnegie Bosch Institute, which is an alliance between CMU’s Tepper School of Business and the German-based Bosch Group, a global supplier of technology and services with 360,000 employees in 60 countries.

The institute was established as an entity within the Tepper School in 1990, through an endowment provided by the Bosch Group. The institute focuses on the improvement of international management through research, education, and collaboration. To that end, the institute sponsors academic research chairs and funds research projects and conferences. Earlier this year it celebrated its 25th anniversary. To honor the occasion, the Bosch Group gave two gifts to CMU:

  • $1 million toward CMU’s Tepper Quadrangle, which is currently under construction on the Pittsburgh campus and will house the business school as well as major university initiatives such as entrepreneurship and technology-enhanced learning.
  • $2.5 million to establish an endowed professorship in Security and Privacy Technologies for CMU’s CyLab, which is a university-wide initiative that establishes public-private partnerships to develop secure computing and communications systems technologies.

Savvides—a research professor in the Electrical and Computer Engineering Department and CMU CyLab—is also the director of the CyLab Biometrics Center. The high-tech space is crowded with computers, robots, and machines and populated with doctoral students working with Savvides to improve biometrics, the science of measuring and identifying people using facial and iris recognition systems. Their team is breaking new ground in the enhancement of existing biometric authentication technologies as well as the creation of revolutionary new ones. Savvides’ research is specifically geared toward improving the use of iris and face recognition in biometric authentication systems. 

Amid the coffee cups, 3D programming books, declassified forensics notes, and facial landmark recognition maps, the researchers don’t just write lines of code; they interact directly with the biometrics machines Savvides has created, constantly working to refine what he has designed, in part, to help find missing persons like Maddie.

Savvides mentions that there are a lot of “Maddies” out there, and he’s right. In the United States alone, more than 80,000 children are currently registered as missing. And according to the National Crime Information Center and the FBI, thousands of people are added to that list every year.

To demonstrate how the machines work, Savvides and his research students huddle around one of the looming bodies of metal and glass, which looks like it was taken straight out of a sci-fi movie. When Savvides moves at a distance from the large sleek robot, the machine immediately picks up on his body and zooms in on his face.

A target that analyzes the eye quickly captures a picture of his iris, which pops up on a connected computer screen. The system shifts from the picture to a split screen of black and green code, matching the image of his iris with the millions of individuals in the computer’s database in real time. Within a few seconds, it’s a match: Marios Savvides.

If the LRI system had existed in 2007 and been widely implemented, such as in airports, it could have helped provide timely information about where Maddie might have been taken.

Eight years ago, the system was originally developed for the Department of Defense for efficient military identification. Placed on a Humvee or a tank, the system could immediately identify any person within scanning distance as soldier, ally, or enemy.

Originally, the LRI system and other face enhancement technology they developed had mostly been tested in labs and on the battlefield. That changed on April 15, 2013:

At 9:30 AM, more than 23,000 runners are at the starting line for the 2013 Boston Marathon. Spectators, including many of the runners’ family and friends, start to line Boylston, a shop-lined street near the end of the marathon’s course. Hours later, three spectators die there and more than 250 runners and bystanders are injured when two bombs detonate.

Later that day, at 5 PM, the FBI releases to the public some low-resolution security camera images of an individual who may be the suspected bomber. But they don’t have the technology to make a positive ID and ask for assistance. Savvides and his team spring to action, and for the next four days attempt to reconstruct images of the suspected bomber. Savvides isn’t sure whether it’s possible; nevertheless, he and his team feverishly download and enhance images, and keep typing out code while land-marking points on faces, and modeling parameters.

Eventually, the team is able to put the low-res images through an algorithm they created. The system brings up an enhanced reconstruction of an 18 year-old man—Dzhokhar Tsarnaev—which they send to the authorities at 2:42 AM, April 19. Several hours later, the identities of the suspects are released after they were being chased during a shootout in Cambridge, Mass. Savvides realizes that one of the suspects is Tsarnaev, which his team identified through the reconstruction of a poor low-res image.

“Holy sh*t!” is his reaction. “I can’t believe we reconstructed this!”

Savvides later demonstrated that he could have honed in on Tsarnaev from a database of a million images; and, in the process, his CyLab Biometrics Center demonstrated that it could solve a real-world problem using code created on the fly. Since the Boston Marathon case, the center has been asked frequently by law-enforcement officials worldwide to help solve crimes. Savvides says the results must be kept confidential, but conclusions about their success rate can be drawn from the ongoing proliferation of requests. “I get multiple pings from law enforcement requesting our help,” he says.

As for Maddie and all the missing children, here’s what Savvides envisions: a future where almost every child who goes missing can be swiftly tracked down. By placing the long-range iris scanner he has invented in key locations such as airports and train stations, children reported abducted, such as Maddie, could be quickly located and law enforcement could then move in.

The scanner’s potential, explains Savvides, lies in the fact that it’s not disruptive. “You say to someone, ‘Hey, can you come up to this machine so we can scan your eye to see if your parents have abducted you?’ It’s too obtrusive. Our system helps to quickly identify people unobtrusively.”

He has secured a patent for the system and believes that within the next decade it could be implemented worldwide. “One of the best parts about it is hearing the police say, ‘You can do this?!’ Like they’re in shock. You can’t beat that feeling,” Savvides says, smiling.

His biometrics research has garnered much attention from, among others, The Atlantic, CNN, Discovery News, PBS NewsHour, NOVA, CBS’s 60 Minutes, Reuters, and MIT Technology Review.In addition, the Edison Awards, an annual competition that honors excellence in innovation and product development, awarded Savvides and his center with its highest honor, Gold, for this technology, which has been implemented in everything from corporate security systems to counterterrorism initiatives.

And, understandably, news about the CyLab Biometrics Center was part of the 25th Carnegie Bosch Institute anniversary celebration held on the CMU campus, which was attended by executives from regional and global industry, faculty, alumni, and business partners.

Aside from all of the hype, Savvides says it’s incredibly gratifying to be involved with such meaningful research: “We’re coming up with new algorithms to solve real-world problems. It’s this incredible feeling to know we are actually making a difference by developing technology that can find an abducted child or helping locate a bad guy.”

Original article posted here.

Related People:

Marios Savvides