Written byDan Verton
How do you detect a person who enters a building through the exit? It may seem like a simple question, but for the Transportation Security Administration and airports across the country it’s a major security challenge.
TSA security checkpoints are designed to ensure that only screened and ticketed passengers have been allowed to enter the secure area of the airport — basically the gate areas where passengers board their flights. But when somebody inadvertently or deliberately enters the secure gate area through the exit — the passageway reserved for arriving passengers as they head to baggage claim or to catch a taxi — security officials have no choice but to force everybody to pass through security a second time.
That process, known as dumping the airport, can cause flight delays and cost millions of dollars.
But researchers from Northeastern University’s College of Engineering, backed by the Department of Homeland Security’s Science and Technology Directorate, have found a way to solve the problem known as “counterflow intruders.” Using advanced video analytics technologies, researchers have been able to not only detect a person trying to enter the secure area through an exit, but to also remember that person’s identity and track him or her throughout the airport.
Known as the Video Analytic Surveillance Transition Project, or VAST, the research program has been underway for the past year at Cleveland Hopkins International Airport. And it’s had a staggering success rate. At one exit, which handles 50,000 people per week, researchers have demonstrated a 99.99 percent detection rate with only five false alarms per week.
“For me, that’s solving the problem,” said John S. Beaty, director of technology programs at Northeastern University. Beaty detailed the program Feb. 11 in Washington, D.C., during a First Responder Technology Demonstration hosted by DHS’ Science and Technology Directorate.
“When we detect somebody going through the exit counterflow, we record that specific video of the event and send it to the control room so that the people who need to make decisions are well informed and they have a picture of the individual … so they can track him through the airport and not have to dump the airport,” Beaty said.
How do they do it?
According to Beaty, one of the major challenges in video analytics is tracking a moving object across multiple, independent security cameras that do not have overlapping fields of view, and often feed their video into separate monitors.
“We call it tag and track,” Beaty said. “You can follow a person through the field of view of one camera, but then you have to re-identify them in the next camera. And that can be a real problem.”
The way Beaty and his researchers solved the re-identification problem was to use three cameras to capture one frontal and two side images of a person attempting to enter the secure area through an exit. They then reconstruct the image of the person in a 360-degree view.
“So now, I don’t care how you present yourself,” Beaty said. “We can identify you. Now I can go from this camera, to that camera, to the next camera, through the field of view and track that person as he progresses from the airport parking lot to the public side of the airport.”
Northeastern is partnering with Siemens to produce a commercial-grade version of the system for use at airports and other security venues, Beaty said. The program is also beginning new partnerships with the Massachusetts Port Authority and Boston Logan International Airport to use their data sets to conduct additional test and further refine the technology.
Video: John S. Beaty describes the VAST research at the Department of Homeland Security, Feb. 11, 2014.