Automated Eyes

ObjectVideo, Reston, Va., has developed a technology capable of doing what people can’t: analyzing video scenes from hundreds of cameras for hours on end without missing a scene.

Over the past two years, U.S. Customs and Border Protection (CBP) has installed ObjectVideo’s VEW software system at ports of entry located along the 7,240-mile largely open borders with Canada and Mexico. Should an unauthorized border crossing occur, the video surveillance system alerts security officers by phone, pager, e-mail or console alarm. The system will also activate alarms if it spots other kinds of suspicious activities: a person loitering in an area, an unattended package, a suspicious parked vehicle.

In early 2003, CBP deployed VEW systems along the Canadian border in the states of Maine and Washington. VEW reduced false alarms so well that CBP expanded the system to cover the rest of the northern U.S. border, the southern border with Mexico, and at seaports and airports that function as international points of entry. In all, the U.S. has 317 ports of entry, 129 along the northern border, 24 on the southern border, with the rest made up of seaports and airports.

Before being commercialized by Object Video, VEW technology traces its roots to the Defense Advanced Research Projects Agency (DARPA), the government research arm. In the late 1980’s, DARPA computer scientists working with artificial intelligence created a science called computer vision to extract and identify unusual objects and events from a video. Computer vision can distinguish between the stationary background and moving objects in a scene. It can tell a person from an animal from a car from a package.

In the wake of Sept. 11, government agencies seeking to raise security levels across the U.S. investigated many new technologies. CBP implemented gamma ray technology to scan trucks for bomb threats and equipped inspectors with radiation sensors capable of detecting dirty bombs.

CBP also wanted to improve the closed circuit video systems that help officers watch the borders. “We can’t build walls along the borders, and we can’t post guards every hundred yards, so we use video surveillance,” says Bill Anthony, director of Washington, D.C. media services for CBP.

Conventional video surveillance proved limited, however. “Up there, wind, rain, and all kinds of wildlife set off the alarms,” Anthony says. “After a while, hundreds of false alarms dull your senses.”

VEW software has made video surveillance practical for border security. The software enables security professionals to create rules for different areas within a camera’s field of view. An officer can draw boxes on a camera’s view of a fence and separate the way the system interprets what happens in the public area outside the fence and the restricted area inside. Conditions can be set for the system to allow vehicles and people to pass through a gate, but to alarm if someone cuts or climbs a fence.

In Blaine, Wash., a major port of entry on the northern border, CBP found that VEW enhances its ability to deal with border incursions related to smuggling operations and the entry of illegal aliens. “They have caught a couple of security breaches there and at other ports of entry,” Anthony says. “It’s a time-saver and an attention-grabber that has made our security officers more productive. It eliminates the tedium of running after false alarms and lets them focus on keeping illegal aliens out — and hopefully terrorists and weapons.”

VEW has also caught the attention of Andrews Air Force Base, home of Air Force One. Andrews has tapped VEW to improve security for flight-line operations such as maintenance, fueling and loading. Most recently, the Jacksonville (Fla.) Port Authority selected ObjectVideo to tighten security along the waterfront and landside of Blount Island, Talleyrand and Dames Point marine terminals.

Currently, ObjectVideo is researching the next phase of computer vision: improved night vision technologies and environmental learning, a technique that aims to eliminate the need to pre-define threats by automating the system to understand an environment with pre-programmed rules.