Hopkins keeps sharp eye on campus

Camera software sees unusual behavior and sends alert to security

September 02, 2005|By Mariana Minaya | By Mariana Minaya,Sun Staff

A line of text pops up on the sleek, flat-panel monitor. Something unusual in the behavior of the figure on the screen has tripped the new security system at the Johns Hopkins University in Homewood.

An operator, sitting in front of four screens receiving images from 32 new security cameras sweeping the campus, double-clicks on the message and watches two images appear on the monitor in front of him.

One image, a few seconds old, shows the figure surrounded by a yellow box the computer has added -- he's walking down an alley. The other is a live image from the security camera of the same alley, now empty.

It's early afternoon, the lighting is good, and the operator can see that the person in the alley is a harmless pedestrian. The operator sits back and waits for the next alert.

As students return to Hopkins next week, they'll find a campus protected by one of the latest developments in surveillance technology.

Once employed mainly in airports and parking garages, it's making its way into use by police and security forces concerned about identifying potential terrorists as well as ordinary criminals.

The $500,000 system can identify 16 different, unusual or suspicious behaviors or objects -- such as a person falling, a lurker in a solitary alley or outside a sorority house, or a suspicious package. It can also identify fast, slow-moving or suddenly stopped vehicles, and it can sense the formation of a crowd.

Once alerted by a text message, the operator can pull up a live image of the area and a history of the images that triggered the alert. He or she can then determine if the situation merits sending a security officer to investigate.

As part of a security shake-up after the deaths of two students last year, Hopkins installed the new cameras, all equipped with behavior recognition software developed by Cernium, a security firm in St. Louis.

In an age of terrorism, some experts see behavior recognition technology as a breakthrough. It has been adopted by parking officials in New York and at George Washington's Mount Vernon plantation in Virginia as well as a variety of airports around the country.

Smithsonian Institution officials are testing it in Washington, and Baltimore police plan to give the system a tryout in neighborhoods along Monument Street, Greenmount Avenue and Park Heights Avenue before the end of the year.

Sophisticated system

Cernium officials said the system is designed to avoid the pitfalls of most closed-circuit TV setups, which typically rely on humans to monitor cameras in real time or on tape recordings that police can pore over after a crime has occurred.

"A human can't effectively monitor a large number of cameras," said Randy Buckner, a technical adviser for Cernium and a professor of psychology, neurobiology and radiology at Washington University in St. Louis. "The software is intended to be used in a preventive rather than reactive way."

Its designers say the Cernium system is more sophisticated than security cameras that merely monitor for movement before issuing alerts.

The software identifies irregular behavior by analyzing camera frames to develop a quick history of an event, such as a person walking in view of the camera.

It calculates how much the brightness and smoothness of an image changes and how quickly a figure changes shape. Aside from spotting a lurker, the software is also capable of noticing abandoned objects, such as suspicious packages, and issue warnings about people who appear to be falling or moving erratically.

It discerns normal from abnormal behavior by adapting to the movement of particular objects. For example, humans normally change shape in a particular way as they walk, with their legs moving rapidly. But a drastic shape change, in which a figure switched from upright to horizontal, would trigger an alert.

The shape change must also be quick to set off the software. For example, it doesn't catch volunteers faking a slow fall. Nor does it raise the alarm when someone bends down to tie his shoe, according to Maurice Garoutte, Cernium's vice president and chief technology officer.

On the other end of the spectrum, the software picks up people who move too slowly -- detecting loiterers.

Helping at Hopkins

The cameras at Hopkins, which watch over public areas with high pedestrian and vehicular traffic, have been useful so far in helping officers intervene in suspicious situations.

Michael Anderson, senior technical consultant for iXP Corp., the company that coordinated the university's security upgrade, recalled one incident in which a security officer was able to approach a man who had stopped and begun to look around in an area of the campus for apparently no reason.

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.