Video algorithms detect people moving against the flow of traffic

It doesn't matter whether you're a terrorist or a two-year-old who's been separated from her parents at the airport: if you cause a disturbance in the flow of airport traffic, you can also cause severe chaos and economic damage.

Regardless of your motivation, moving the wrong way through a security checkpoint is treated as a threat by the Transportation Security Administration. Breaches of this sort can cause airports to temporarily shut down, flight delays, and result in the loss of millions of dollars.

"You can imagine the cost and the angst that occurs if you don't deal with this," said Michael Silevitch, director of Northeastern's Awareness and Localization of Explosives-Related Threats Center, a Department of Homeland Security Center of Excellence. "So we dealt with it."

ALERT's Video Analytics Surveillance Transition, or VAST, team is working to help minimize the damage. In collaboration with the Cleveland Hopkins International Airport, the TSA, and Siemens Corporate Research the group—which includes researchers at Northeastern, Boston University, and Rensselaer Polytechnic Institute—aims to help TSA officers make better use of the copious video data at their disposal. Their research was recently featured in an article on the website fedscoop.com, which praised the work for its ability to "solve major security problems for airports."

"The data we want is usually just a few pixels in a few frames within gigabytes of data," explained electrical and computer engineering professor Octavia Camps, who leads VAST's Northeastern arm.

Cleveland Hopkins International Airport has 350 security cameras installed throughout its terminals. At a large international airport such as Boston's Logan airport, that number is more like 1,500. Currently, the standard way to investigate the footage collected on those cameras is with the human eye. VAST is automating the process.

In a project called "in-the-exit," three separate algorithms are being optimized to detect isolated moments in the footage when people are moving against the flow of traffic. This significantly reduces the amount of video that TSOs need to review, allowing them to focus more on the type of breach they're dealing with.

"The question is can we identify breaches of that type, flag the security officer, and flag the breach before the person actually gets back into the secure area," said Silevitch.

The program, which has so far been operating in a research capacity at Cleveland Hopkins, has a 99.9 percent detection rate with an average of five false alarms per week.

In-the-exit has been tested since January 2013 and will soon be incorporated into everyday operations at the Cleveland airport. But it is just the first project to be optimized with VAST. The team's next step is to focus on tracking individuals as they move throughout the airport, passing in front of various cameras, which each maintain a different field of view. Using features such as the color and texture of a person's clothing and the person's shape and size, these new algorithms will be able to automatically identify the same person at various points throughout the facility.

Provided by Northeastern University