Preventing Campus Crime With Cameras

• By Michael Fickes
• 08/01/10

Around 3:00 A.M. not long ago, on a university campus in Baltimore, a woman walking a dog turned into an alley and spotted three men in hoods racing toward her. Terrified, she jerked the dog’s leash and started running.

Suddenly a security vehicle with two university security officers wheeled off the main street into the alley. The vehicle’s headlights trained on the hooded men, who turned and fled.

Lucky break for the woman? Yes and no. Luckily, the woman was walking near the Johns Hopkins University Homewood campus, which had deployed intelligent video surveillance technology.

The new technology, however, doesn’t need luck. Intelligent video systems can identify certain human behaviors, including one person or a group of people converging on another person or group.

In this case, a video camera trained on the alley picked up three people converging on a lone person. Software operating in the video system reacted by sending an alarm to the University’s security center and the video to a large monitor in the center, where a security officer studied the video. Seeing a developing crime, he instantly dispatched a nearby security vehicle.

When the security vehicle arrived in the alley, the hooded men fled.
 
The video technology that helped prevent the crime is known generically as video analytics. It is a powerful technology, which, in the hands of a trained security department, can literally prevent crimes — not all crimes, of course, but many.

Crimes on campus — robberies, assaults, burglaries, motor vehicle thefts, arson, and worse — damage an institution’s reputation, recruiting efforts, and retention capability, said Lawrence Consalvos, senior vice president and general manager with iXP Corporation, a security consulting firm based in Cranberry, NJ.

A technology like video analytics that can help prevent crime can also protect reputations, recruiting activities, and retention capabilities.

What Are Video Analytics?
“Studies show that people cannot effectively monitor video screens,” said Consalvos. “One person can monitor video from 20 cameras for about 20 minutes. After that, he or she loses focus.”

Intelligent video technology can monitor effectively. Unlike people, technology never loses focus. A video analytics application sees every single person or group of people walking past every camera or loitering somewhere within a camera’s field of view. It sees groups of people converging. It sees cars and other moving objects as well as abandoned and stationary objects.

Better yet, it is possible to program video analytics applications to pay special attention to behaviors that might indicate a security or safety problem and to notify a human being, who will make a judgment about whether or not the behavior in question needs to be investigated further.

Video Analytics at Johns Hopkins
The Johns Hopkins University Homewood campus in the City of Baltimore spans 140 acres with 62 buildings. Approximately 4,500 undergraduate students and nearly 2,000 graduate students attend classes on the campus, and about 6,000 faculty and staff work there.

As if a campus the size of Homewood isn’t challenging enough for a security department, Baltimore City is a relatively high-crime city. In 2009, the CQ Press released its annual City Crime Rankings 2009-2010: Crime in Metropolitan America, which ranks U.S. cities with at least 75,000 residents according to statistics released by the Uniform Crime Reporting Program of the FBI. The study ranked Baltimore’s crime rate 13th highest in the country.

Edmund G. Skrodzki, executive director, Campus Safety and Security for the Homewood campus, came to Johns Hopkins in 2005 following a 28-year career in law enforcement, including 22 years with the U.S. Secret Service. Since his arrival, he has labored to make the Homewood campus an island of security and safety within Baltimore. Intelligent video plays a key role in that effort. 

Under Skrodzki’s direction, iXP Corporation began installing video surveillance cameras paired with video analytics on the Johns Hopkins University Homewood campus back in 2005. Today, 171 of 193 video surveillance cameras on campus send video to a video analytics system.

The Homewood campus’ video analytics system works on a bank of video servers located adjacent to the Homewood Communication Center, the main campus security office, and a police/911 dispatch center capable of two-way radio communication with campus police as well as Baltimore police, fire, and emergency medical services.

The specially programmed analytics application searches all of the video for 10 specific actions, images, or situations. These include:

• A single person walking
• A loitering person
• A group of people
• Converging people
• A single vehicle
• Multiple vehicles
• An object left behind
• An object removed
• An object in motion
• Any movement within an area defined in a camera’s field of view

A single person walking across a quad at noon on a warm spring day probably isn’t something to worry about. Then again, a single person wandering around alone on campus in the middle of the night might be a security risk. The system can be set to alarm on a single person walking, a loitering person, multiple people, and so on when a camera captures those video images at an unusual time of day or in an area where people aren’t supposed to be.

When the analytics system sees one or more of the programmed behaviors, it notifies a security “specialist” in the Homewood Communication Center. “The lag time between an event and the arrival of the…alert is less than two seconds,” noted Skrodzki. “The system records the alert time and the time that the specialist acknowledges the alert. Management constantly reviews these times with the goal of minimizing any human delay in the process.”

The system currently produces about four alerts per minute, continued Skrodzki.

Specialists, trained to recognize situations that aren’t quite right, review each video-alert clip and decide whether or not to investigate further.

Results
When Skrodzki arrived at Hopkins in 2005, bike theft was a common campus crime. Each year, 30 or more bikes locked to racks supplied by the University disappeared.

Today, video analytics have virtually eliminated bike theft.

In one typical incident, a thief appeared next to a bike rack and cut the lock with a cutter. Instantly, the analytics system alerted the specialist, who recognized the crime and dispatched the closest security officer. Within 11 seconds the thief had hopped onto the bike. But before he started pedaling, a Hopkins security officer stopped him. Eleven seconds isn’t fast enough to steal a bicycle on the Homewood campus.

Bike thefts have plunged. In each of the last two years, only five bicycles have been stolen. Most of those thefts occurred outside of a camera’s field of view.

Bicycle thefts are one thing, but serious crime has dropped, too. In 2003, prior to the installation of the video analytics system, Hopkins’ Homewood campus reported 52 serious crimes: five aggravated assaults, two incidents of arson, 17 burglaries, two forcible sex offenses, 18 motor vehicle thefts, and eight robberies.

In 2009, the Homewood campus reported a total of three serious crimes: a robbery and two stolen cars.

Of course, the reduction in crime at the Johns Hopkins Homewood campus isn’t entirely attributable to video analytics, but it is certainly an important part of the story.

Skrodzki noted that the Homewood campus has not only added video analytics, but also increased the size of its security staff, modernized its equipment, and integrated a number of new technologies into the system since 2005.

The common theme to all of those changes is Skrodzki’s goal of preventing crime on campus. As part of the overall system, then, video analytics has at least earned some of the credit for the good results at Johns Hopkins.