Aug 1, 2002 12:00 PM
By CORRINA STELLITANO
Last year, 53 million tourists entered America's casinos hoping to catch a stroke of good fortune. Lurking among them were sinister professionals who refuse to depend on mere luck.
Cheaters who tour the nation's casinos swindle millions of dollars annually. In the past, the task of identifying and apprehending these culprits was as difficult as pulling three bars on a slot machine. Surveillance personnel who observed a suspicious person on the casino floor would pore through thousands of mug shots to find a match. Hours or even days later, they might see the face of their suspect staring back from the pages. By then, the cheater would be somewhere else making another quick buck.
Today, face-recognition technology offers casino surveillance teams the capability to compare a suspect's face with a database of thousands of pictures in only seconds.
As president of Biometrica Systems, Bob Schmidt began offering his facial recognition solution "Visual Casino" in 1998. Today, Viisage Technology Inc., Littleton, Mass., which acquired Biometrica in March, provides face-recognition software and extensive databases to more than 150 casinos.
While not every industry is suited for the technology, Schmidt, who now serves as Viisage's executive vice-president, says casinos are an ideal application. "There are three things you need for facial recognition to be successful," he explains. "First of all, you need some kind of financial or security justification. You need a database and you need cameras. The casino market fits all three.
"They had a huge need because they lost millions of dollars every year to cheats and [card] counters," Schmidt continues. "They had a database of known cheaters and card counters in the industry, and of course, they had cameras. Just walk into a casino and look up at the ceiling. There are hundreds."
TAKE YOUR BEST SHOT
To supply face recognition capability to casinos, Viisage's computers are attached to the casino's video system. Camera operators can direct any camera to capture an image of a suspect's face, almost instantly entering the photo into the face-recognition search engine. Most often, surveillance personnel will use the system to identify a suspicious person.
In the 78,000-square foot Sands Hotel and Casino in Atlantic City, more than 1,000 cameras scan for suspicious activity. Charlie Guenther, executive director of surveillance for the Sands, explains: "Maybe you have a person on the casino floor who is looking or acting suspiciously, or cheating; or someone recognizes him as someone who has been previously evicted, ejected or arrested. You would zoom a camera in, get the camera as close to a straight shot on the face as you can, hit a button and you're done. The computer does all the rest in seconds."
Face-recognition software translates the characteristics of the face into a unique set of numbers called the eigenface. Using either a still image or a shot from live video, the software calculates 128 different measurements of the face, from hairline to chin and ear-to-ear. The result of the calculations is "a unique identifier for [the mapped face] at that point in time," Schmidt says.
The result is compared with other faces in one or more databases, and probable matches are identified ¡ª accompanied by physical characteristics, aliases, and other helpful information. The clearer the picture, the quicker the matches are delivered, says Will Bennett, surveillance director at San Diego's Barona Casino.
"The old computer adage: 'What you get out of a program is only as good as what you put into a program,' is the bottom line," Bennett says. "In a casino, one of the key elements is physically positioning certain cameras with facial recognition in mind. Then [the technology] works wonderfully."
Poor-quality photos are not impossible to match, however. George Busch, director of internal controls for the Mystic Lake Casino in Minneapolis, tells of the capture of a fugitive wanted by the Federal government using a five-year-old photograph. "It so happened we had a Federal agent in our surveillance room and the suspect was on the floor at a gaming contest," Busch recalls. "The truth of the matter was [the software identified] him, even though his facial features had changed."
A DIGITAL PHOTO ALBUM
Viisage users can subscribe to a nationwide database operated by Casino Visual Identification, and can use Viisage software to create their own in-house database. These databases include individuals who have been caught cheating or using other organized gaming deceptions, and individuals who have been ejected for offenses such as drunk and disorderly conduct.
Nationwide databases also serve to network many casinos together, thus allowing quick notification when a cheater is in the region. Bill Doolin, surveillance manager for Harrah's Joliet Casino in Chicago, first worked with face-recognition software when it was installed at the casino in April. He praises the immediacy of the system.
"As other properties which belong to that network catch somebody, they send a report that goes to all of the other casinos in the network," Doolin says. "Within an hour or two of them catching someone, maybe in Nevada or in Atlantic City, all of the casinos in that network will have that information available to them."
THE HUMAN COMPONENT
Despite the technological aspect of face-recognition systems and their extensive databases, human involvement is essential to the system's success.
"There's no surveillance system today that I would use and let the computer make a decision. The computer can get you to the question of 'Is this the same person?' but a human has to be the final decision-maker in a surveillance application," Schmidt says.
New face-recognition cameras, such as Biometrica's BiometriCam, include built-in recognition software and can be used to continuously scan an entrance or exit. The faces of those entering the casino can be compared to a smaller personalized database and alert surveillance departments of suspicious entries, thus allowing final visual verification.
When training casino surveillance teams, Viisage trainers emphasize the importance of the human opinion as superior to the recommendation of the computer.
"People today will defer to a computer," Schmidt explains. "We were terrified that someone would look at two pictures and say, 'It sure doesn't look like the same person to me, but the computer says it's 98 percent, so it must be,' and then take some action. We want them to be conservative. We do not want them to be too aggressive in going after people, because you're messing with customers."
As a final safeguard, security officials on the casino floor are equipped with handheld wireless units which allow them to view the photographs before taking action.
If a human is still necessary to complete the match, how helpful is the technology? Casino surveillance professionals describe face-recognition software as a tool which aids humans, but doesn't replace them.
"A new staff member doesn't have to come in and start a learning curve that takes months to complete," Bennett says. "On the first day on the job, they can see what they consider suspicious behavior and ask our database if this is someone who we have had contact with before and get an instant report. It gives them immediate access to the history of an individual rather than having had to live long enough to have personal history with someone."
The system's processing speed surpasses even skilled law enforcement professionals who make their careers in casinos. "Surveillance people have a talent for names and faces because that's our job," Guenther says. "I've been doing this job for 22 years, strictly surveillance, and I'm not ashamed to tell you this computer is quicker and faster than I could ever imagine being."
Guenther witnessed the first test of the software at Atlantic City's Trump Marina while serving as director of surveillance. "When we installed it back in February of '98, we had a large group of cheaters, and when we went to identify the seven members of the team, the computer was able to identify these people much quicker than all of us. I realized [then] that this would be a system that would help everybody in identifying faces in a quick and extremely accurate way."
Corrina Stellitano is a Fairhope, Ala.-based writer and regular contributor to Access Control & Security Systems.
VERIFICATION VS. IDENTIFICATION
In addition to casinos, Viisage Technology Inc., Littleton, Mass., also provides face recognition capabilities to law enforcement agencies, airports and to companies requiring specialized access control. The search capabilities required by each application differ slightly, but the same technology is used. The primary difference in search systems is described as verification versus identification.
"That's a very important concept in facial recognition," says Bob Schmidt, Viisage's executive vice-president. "In one case, (verification), ¡ the computer is really doing nothing more than comparing two things. And that's a fairly simple thing for a computer to do." When an employee seeks entry to his office building by presenting a key card and then showing his face to the camera, the face-recognition system is engaging in verification ¡ª and confirming that the employee's picture on file matches the information and face presented.
In law enforcement and casino applications, the computer must search for matches for one face among many others. "When you're doing an identification, you're trying to pick somebody out of a database of 5,000 people and that requires a much more sophisticated search capability, because the computer has to actually search based on facial characteristics," Schmidt says.