, securing transit in-frastructure has been uppermost in the transit industry’s collective mind. Deploying video surveillance systems or upgrading those already in place have become key elements of transit security and safety programs being deployed, which transit systems have been able to do with the help of U.S. Department of Homeland Security (DHS) grants. While building on video surveillance programs by adding additional cameras will help bolster these programs, figuring out how to manage the data from the cameras is another story.
The San Francisco Municipal Transportation Agency (Muni) is one of several transit operations using DHS grant monies to upgrade its program. The agency recently signed a $2 million, five-year contract with Behavioral Recognition Systems Inc. (BRS Labs), developer of video behavioral recognition software, with U.S. offices in Houston and Washington, D.C. The company has developed advanced, intelligent software that uses behavioral recognition technology to learn — on its own — about the environment and objects it observes in each camera’s field of view.
Per the Muni contract, BRS Labs will install the software, called AIsight, in 12 stations, with 22 surveillance cameras in each location, according to Muni spokesperson Paul Rose. Installation of the video surveillance cameras is expected to be completed by the end of 2013.
“This technology gives us another tool to identify abnormal behavior within our system to ensure our employees and passengers are as safe as possible,” he says.
“Since 9/11 really, there’s been an explosion in the utilization and planning to expand video surveillance technology,” says BRS Labs’ CEO John Frazzini, adding that not only is this technology growth being felt in the U.S., but globally as well. “You look at the 2004 Madrid train bombings, you look at what happened in London, and so points of compromise in Europe have been in transit systems.”
These events have led the U.S. to require more hardening, or more sophisticated security applications for transit, which includes the use of video surveillance technology. But, equipping rail stations and other key transit infrastructure areas with video surveillance cameras leads to the question of what to do with the information collected by these camera systems.
“If you look at it from a common sense perspective, if you’re deploying hundreds of thousands of video surveillance cameras, currently there’s no mechanism for humans to monitor those cameras 24 hours a day, seven days a week,” Frazzini says. “If you put up a camera system of 5,000 cameras, what proactive value are you going to get by hiring several security guards, for example, to monitor those?”
This leads to a lot of head scratching by organizations, according to Frazzini, wondering how they are going to deal with all of the data.
“We solve that problem,” he says. “We set forth the technology that analyzes, through a very artificial intelligence framework, the video surveillance data collected by these camera systems being deployed around the world.”
The software processes the information being collected by those cameras, which Frazzini likens to being the “eyeballs of the video surveillance system,” and the company provides the brain.
The way the system works is, once the system is installed in the video camera, it takes a few weeks to form its initial understanding of each camera’s field of view so it can observe things over the course of time.
“If you were to take a look at a video surveillance camera in the first 15 minutes, you would have to use your gut to determine whether what you are looking at is normal or not, because you don’t know,” he says. “What happens on a Friday at 3:00 p.m. is different than what happens on a Sunday at 9:00 a.m.”
The system needs a few weeks for it to gain awareness and an understanding of the field of view.
After the system has formed the initial understanding, it creates a memory set, which is basically the system’s memory. Once the baseline understanding is formed, if the system observes behavioral activity that deviates from that established pattern, it issues an alert in the form of a five- to ten-second alert clip. This is a video clip that is then presented through a command and control system to the customer for further review.
“Our system produces the video surveillance intelligence that operators now look at to determine whether a response is required,” Frazzini says.
While the science behind the software is very complex, he adds, the utilization is very straightforward, as transit systems need only plug it into the existing camera environment and the system starts analyzing the data and learning from it immediately.
After a few weeks, the alerts are presented in a format similar to a YouTube environment, where an operator just has to click on a video clip and make a determination as to what the response is. Video clips can be sent to smart phones and disseminated through email like any other video clip.
This software also reduces the need for camera maintenance. If, for example, a camera turns off, breaks, or is moved, the system will alert the operators, notifying them of a substantial field of view change for each camera.
“We’ve automated the maintenance process in these large, broad video surveillance networks where you don’t have to do system health checks,” Frazzini says. “It’s one of the associated benefits of using our sophisticated software. This associated benefit of the technology saves a substantial amount of time, energy and resources in just using the camera infrastructure by using the software’s ability to identify when something just isn’t right.”