AT&T Technology Sponsorlogo

Finding a Face in the Crowd

A security guard looks at the images on the screen in front of him: a video capture of a clean-shaven man with close-cropped hair next to a mugshot of a convicted American drug dealer with a beard and shaggy long hair. They look different, but the guard confirms they are definitely the same man. In minutes, the positively identified man is apprehended.

Melanie Basich 2012 Headshot

As a group of passengers exits a plane at Toronto International Airport, a security camera scans the face of each person deplaning. The camera scrutinizes a short, gray-haired lady, then a tall blonde. Finding nothing, the camera focuses in on a handsome clean-shaven man with short, cropped hair. Just like it did with the two women, a system locates the man's eyes and nose and then measures the distances between key features on his face. Special software compares his image against a database of thousands of mugshots and, unlike with the two women, a match is found. A warning message appears on a screen alerting security that a match has been made. All in a matter of seconds.

A security guard looks at the images on the screen in front of him: a video capture of a clean-shaven man with close-cropped hair next to a mugshot of a convicted American drug dealer with a beard and shaggy long hair. They look different, but the guard confirms they are definitely the same man. In minutes, the positively identified man is apprehended.

If facial recognition works this well with known drug dealers, imagine what it could do with known terrorists entering an airport.

How it Works

Facial recognition software scans a picture of a face-from a photograph or a video freezeframe-looking for identifiable features. These features are then encoded into a string of numbers that represents that person's image. This set of numbers is compared against sets of numbers in a database until similar images are found. The computer will alert the officer conducting the search of a match or near match.

The officer can then view the images the computer has identified as possible matches and verify a match. Many prefer facial recognition because a human can double-check a computer match, unlike fingerprints or retinal scans, which cannot be verified with the naked eye.

Because it stores simple lines of code instead of memory-intensive images, facial recognition software can search through millions of people for a match and get results in seconds. This can even be done from a laptop or handheld while on the street.

The Software

All facial recognition products work in pretty much the same way, but there is more than one way to map a person's face. Each company has a slightly different approach and different terminology.

Face-It Argus from Identix-which recently merged with Visionics-measures  relative distances between the peaks and valleys of the face, using about 80 points to develop a unique image, converted into a string of numbers. Because the features in the center of the face are the most stable, Face-It Argus focuses on this central region, according to Frances Zelazny, director of corporate communications at Identix.

With a different approach, ID-2000 from Imagis Industries uses "light reflectants" to recognize features across the entire face, explains Imagis President and CEO Iain Drummond.

"ID-2000 finds the two eyes and the tip of the nose. It then looks for about 200 features in the face. A feature is a curve, such as the hollow of the cheek, the shape of the cheekbone, the curvature of the eye socket, that type of thing. What we do is we detect it by changes in light reflectants, which is how human beings detect facial shapes. For instance, you see a difference in shading as you look at the hollow of a cheek.

"Having found these 200 features, we then transform them mathematically into a digital string that's about 400 bytes long. And that string becomes your unique facial signature."

Viisage uses yet another formula for its two facial recognition products. Both FaceFinder, designed to identify faces in streaming video, and FaceExplorer, most often used for still photographs, are based on what the company calls the "eigenface" template. The template consists of 128 different facial measurements between the chin and the middle of the forehead and from ear to ear. Michael Mazzu, vice president and general manager of public safety for Viisage, calls it "masking off the area."

Mazzu explains, "A template is just a series of 128 numbers. That template of 128 measurements, each one of which is basically a number, is compared to other sets of these numbers in the database."


With so many different makers of facial recognition software, one might wonder what happens when different police departments choose different software.

Every police agency in Alameda County, Calif., uses facial recognition software from the same company, which makes sharing information easier. But it is not necessary for police departments to use exactly the same technology in order to trade mugshot information.

Many facial recognition systems are also sold with booking systems and CADS software. But for those agencies that only want to buy the facial recognition software, this set-up can work equally as well.

Drummond of Imagis Industries says facial recognition software is made to work with any booking system or database, even when agencies have the software  customized to their specifications, a service offered by most companies.

"If the mugshots are in a different database, then as long as we know where those mugshots are in that database, we can point the facial recognition system at the mugshots and it will use that database rather than our own. It can use both companies' databases at the same time."

And even if departments have different facial recognition systems developed by different companies, they can search each other's databases of images.
According to Viisage's Mazzu, "Although all the photos are done in a standard format, we're not going to utilize another company's coding of images. But there's no need to if you have access to the images themselves."[PAGEBREAK]

Practical Applications

Once a department has decided on a system and successfully installed it, the agency can use the software to help solve multiple problems that plague police officers.

Many people with a history of arrests give false names when arrested, hoping to fool police and get a lesser sentence as a first offender. But searching for a person based on facial features will bring up the correct person, no matter what name is given.

Sgt. Guy Dove of the Huntington Beach (Calif.) Police Department has had great success using facial recognition technology to identify people who refuse to give their real names during booking.

"A lot of our arrests are recurring. They just keep coming back over and over. So facial recognition is very beneficial to us."

Officers can also use facial recognition to identify people who refuse to give any name at all. Not to mention its use in investigating forgeries, fraud, and identity theft.

But facial recognition isn't just used to find criminals or dangerous people. The technology also helps find missing and exploited persons.

Tampa, Fla., and Virginia Beach, Va., now keep an eye on public places with cameras hooked up to Identix facial recognition software. These cities' programs are designed to catch crooks as well as find missing persons by matching images on the camera to databases in local police departments.

In another use for facial recognition, Imagis Technologies was contracted by the U.K. National Crime Squad to develop a system for Interpol to identify missing children being used for child pornography.

This system has started in the U.K., but will eventually search for missing children globally. By including the ability to recognize common backgrounds in pedophilic pictures, authorities hope to use Imagis' program to link various children appearing in photos to the same photographer...and catch him.

Future of Facial Recognition

Unfortunately, criminals are forever looking to evade the authorities. Many people in the industry believe 3D imaging will be the next incarnation of this facial recognition to help police stay ahead of the game.

With the current 2D face recognition technology, if a person's face is not fully in view in a video or picture, his face cannot be imaged. With 3D imaging, a person's profile could be mapped and matched against a database as easily as a full front view.  Researchers are also looking to improve the ability to recognize faces in different lighting situations.

FaceVision by GeoMetrix is already creating 3D imaging using two still cameras. Other companies are looking at using this technology as well, but most are still in the development stages.

3D imaging with a 3D camera is not widely in use yet because of the money and logistics involved in upgrading to such a different system.

"Police have been taking mugshots for a hundred years now," says Imagis' Drummond, "so they already have the databases of criminals to match against. The mugshot databases all have 2D photographs. If you're going to use 3D, you really have to match a 3D image against a 3D image. You need to use 3D cameras and 3D cameras are much more expensive. We've evaluated these cameras, but until it's clear that police are wiling to invest about five to six times the cost of a standard camera, 3D cameras are generally not useable."

Viisage's Mazzu has similar sentiments. "We're already doing things in the development of 3D technology, but the actual devices aren't commonplace by any means yet. That will come eventually, I believe, in the future. There will be much more specialized applications first."

People in the industry expect the 2D form of facial recognition currently in use won't be going away anytime soon. And neither will other types of biometric identification. Mazzu predicts facial recognition "will become a tool used even more than fingerprints. You're not going to do away with fingerprints, but does that mean you're not going to use facial recognition? No, you'll use them both."

Privacy Concerns

People want the country to be secure, but they don't want to feel like Big Brother is always watching them, violating their rights to privacy. Identix President and CEO Joseph J. Atick recognizes the importance of this concern to the public.

"As our customer list grows, and more agencies adopt facial recognition to enhance public safety, Identix continues to work hard to ensure that all installations adhere to established responsible use guidelines."

Identix follows specific privacy principles regarding public areas where its Face-It software is in use. Signage informing the public that the system is in use is posted in corresponding areas when possible.

Most people are aware that security cameras in airports are now often equipped with face recognition systems. But now they are popping up in casinos and shopping malls, among other public places. There was a public outcry when it was discovered that facial recognition was being used to scan every person at Super Bowl XXXV in 2001.

According to Arthur Zwern of 3D imaging company GeoMetrix, "The industry is moving toward finding ways to protect privacy while also finding ways to solve crimes."

But public perception is not always the defining factor in the way public safety runs its business.

"Is the Super Bowl the best place to use facial recognition to find criminals?" asks Zwern. "Probably not. Is an airport a good place to do that? Well, after 9-11 I'm pretty sure most Americans would say, 'Yes, take my picture in the airport all you want if it's going to keep me from getting blown up in an airplane.'"

But according to industry leaders, precautions are taken. Faces that do not bring up a possible match in a database are immediately erased and discarded. This is done both to preserve privacy and to make room for information on true threats.

For More Information


Identix Incorporated

Imagis Technologies Inc.

Viisage Technology Inc.

About the Author
Melanie Basich 2012 Headshot
Managing Editor
View Bio