In the wake of the Christmas "underwear bomber" attempt in Nigeria, airports and other critical security facilities are adding increasing numbers of cameras to boost security. ((iStock))

In 1989, Montreal's École Polytechnique was the site of the tragic killing of 14 women by an enraged gunman. Security for the college's three-building campus has since been beefed up with hundreds of high-end video surveillance cameras, but the installation of so many cameras created a new problem.

"The security director told us last year if another crazy person started running down the hallways with a gun, he wouldn't be able to track him with traditional video surveillance monitoring systems," says Christian Laforte, co-founder and president of Montreal-based Feeling Software, a 3D software design firm.

Traditional video systems in facilities with hundreds of cameras are difficult to use for tracking suspects, explains Laforte, because humans have trouble keeping track of where cameras are located and transitioning from one to the next to follow a moving subject.

The problem is particularly acute in airports, where fears about armed terrorists run high. Larger ones such as Toronto's Pearson International Airport have almost 1,000 cameras in place, and transfer hubs such as New York's La Guardia have thousands, he says.

To use them effectively, security guards first need to memorize where all the cameras are and what they view, an exercise that can take months for new guards. Watching for potential incidents on monitors that display all the video feeds from different areas simultaneously is also difficult.

"A guard typically watches six screens, and each one can have 16 to 32 camera views. They can only watch for about 20 minutes before they burn out," says Rene Beaulieu, president of Securaglobe, a security firm based in London, Ont.

In the wake of the Christmasunderwear bomber  attempt in Nigeria, airports and other critical security facilities are adding increasing numbers of cameras to boost security. "But then guards have less time to view each camera and it dilutes the opportunity to recognize something abnormal," says Robert Smallback, CTO at the RCSC Group, a security firm based in Cape Coral, Fla. "Knowing which camera to follow becomes an impossible task."

If an incident occurs and suspects are running through a facility, guards need to select a series of cameras sequentially from a list, redirecting from one camera to the next nearest one to follow their movements.

"It takes a few seconds to react and select each camera, so even guards who are very good at this will soon lose sight of the suspect," says Laforte.

New technology solves old problems

Technology companies are developing innovative solutions to deal with the growing problem of security camera information overload.

Montreal's École Polytechnique tackled its problem by adopting Feeling Software's new 3D video surveillance management system, says Laforte, who has high-end 3D credentials. In 2002 he won an Academy Award for 3D design as a key team member of Alias-Wavefront, a Canadian company that developed 3D special effects software used extensively by Hollywood movie-makers today. He started up his own company in 2005.

'Airports are drinking from a fire hose when it comes to video.' — Stephen Russell, 3VR Security

Feeling's system tackles video security the problem by creating a 3D model of the interior and exterior of a facility, complete with cameras, so guards can follow a moving suspect visually in a natural way. It takes video feeds from all the cameras and superimposes them on a scale model of the building. The result is a real-time, almost God-like view of the entire area that can be easily manipulated on a computer screen.

"When you want to follow someone, you just click in their direction, and the system automatically computes the next logical camera. It looks like projectors are being rotated and positioned on-screen as they follow the real cameras," says Laforte.

Finding the needle in the haystack

Other companies are taking on other aspects of the problem. A San Francisco-based technology company, 3VR Security, has developed a video search engine to handle the vast amounts of video surveillance data being churned out and stored. It makes isolating video of a specific series of events much easier.

"Airports are drinking from a fire hose when it comes to video," says CEO Stephen Russell.

The issue was underscored in a security breach at Newark Liberty International Airport in January when a man crossed into a prohibited boarding area, then disappeared into the crowd. Investigators shut down the airport for hours to track him down — and learned later he was not a terrorist but a Chinese student kissing his girlfriend goodbye.

"With 3VR's system, officials would have been able to search video stored around the time the incident occurred, be alerted the next time he walked by a camera, and sent security right away to interrogate him," says Russell.

He adds that 3VR's system is already being used at Evansville Regional Airport in Indiana. Russell describes it as a sort of "Google for surveillance," and security teams can use it to search video footage not just for faces, but also for specific objects, colours and shapes.

Face recognition systems that can match people's faces on video images to external criminal databases are often inaccurate, but 3VR's system works on a different principle, he explains.

"By definition, Google generates many false positives for every correct result, yet it's a wonderful tool, because it's better than going through billions of web pages," Russell says. "Our system takes unmanageable volumes of video, and if you're looking for a face or a dress, you can filter down millions of hours of video to a small enough subset for humans to look through in minutes."

Strength in numbers

Smarter video management systems that can tie together multiple video surveillance systems are also being developed, says Peter Wilenius, vice-president at March Networks, an Ottawa-based video systems provider.

Many parties often own the cameras in different areas of an airport, he explains. "In Canada, CATSA (Canadian Air Transport Security Authority) owns the cameras around security checkpoints, the airport authority may own those within the terminal, and airlines and even food courts may own the ones in their areas. And there are multiple stakeholders like the RCMP and customs agencies that need to share video evidence in a crisis."

By tying these systems together, security teams can get a more complete picture of what's happening in the facility they're safeguarding.

New IP-based video systems are also armed with intelligence to detect whether cameras are offline or not functioning properly, and video analytics that can automatically detect anomalies such as motion in restricted areas to alert guards about a potential breach, he says.

New capabilities are being developed to thwart people who steal or otherwise acquire access control cards to restricted areas, passports and other identity documentation, he says.

"There's an increasing need for video management systems that tie video surveillance to other systems like access control and passport readers. With our system, we're able to interrogate the access control database and associate it to an incident in our video database to determine if a face matches the access control card," says Wilenius, adding that a March system was recently implemented at the Melbourne airport in Australia.

While the use of video analytics and other computer-based video management systems are useful in particular areas and environments, Smallback says there are limitations to all this technological wizardry.

Computers can't be used to detect abnormal human behaviour — which is the most critical security piece in airports.

"One country's normal behaviour may be abnormal in another. In an international airport like Vancouver's, you have people with different cultural backgrounds coming from all over the world. Abnormal conditions like movement in a restricted area can be identified by computers, but abnormal human behaviour can't be defined with computer logic," says Smallback.