October 26, 2010 By Russell Nichols
At a strategic surveillance hub in East Orange, N.J., caution alerts triggered by sensors strategically located across the city flash on flat-screen TVs faster than the speed of crime.
When that happens, operators scan live video feeds to search for signs of suspicious activity. If it looks like a crime in progress, operators can dispatch the nearest available patrol units to the scene with the touch of a button.
Blink and you might miss the action. At the Real-Time Crime Prevention Center, the sensors, or “smart cameras,” are among the latest high-tech tools in the East Orange Police Department’s (EOPD) arsenal of gadgets aimed to stop crimes before they occur. The devices slash response times from minutes to mere seconds — a critical difference in a city plagued by drugs, gangs and shootings, said Jose Cordero, East Orange’s police director. “It’s no different from what we do on the street,” Cordero said, “but now we have a system in place that can look at our data and turn information into intelligence in real time.”
For the past few years, the EOPD has been building its $1.4 million state-of-the-art crime prevention network, which Cordero said includes gunshot detection systems, surveillance cameras, wireless computers and software that breaks down crime data instantly. In 2009, the department acquired 10 smart cameras for $195,000 to incorporate into the grid as a force multiplier. But in many ways, the sensor technology is so new, advanced and untested that some refuse to call it a sure-fire, crime-fighting solution.
“The notion that technology will pick out criminals and turn itself on at the right moment and enable the arrests that need to be made is a nice idea,” said Dennis Kenney, a criminal justice professor at New York’s John Jay College of Criminal Justice, “but there is no reason to believe it will work.”
In the face of skeptics and privacy advocates crying Big Brother, local law enforcement officials fired back with striking statistics: Since 2003, violent crime in East Orange has dropped more than two-thirds.
“We didn’t get to where we are by merely reacting to crime,” Cordero said. “If we get information that something may happen and we want to prevent future incidents, we use virtual resources.”
The concept of stopping crime ahead of time has been around for ages, usually spilling from the pages of science fiction stories. But with the rise of forecasting technology in recent years, more government agencies want to look to the future to stay ahead of the curve.
Since 2006, for instance, the Memphis (Tenn.) Police Department has been using software to evaluate incident patterns throughout the city and forecast criminal hot spots. With IBM predictive analytics technology, the department claims to have reduced crime by 31 percent. And in Florida, where 85,000 kids flood the state’s juvenile justice system every year, the Department of Juvenile Justice also implemented predictive tools that analyze past offense history, demographics and other factors to determine which kids have a higher chance of recidivism.
But the sensors used in East Orange see things differently: Rather than peering into the past, these smart cameras catch potential criminals red-handed in real time. Developed by Australia-based company DigiSensory Technologies, the Avista Smart Imaging Sensor system consists of the 3.2-megapixel smart IP camera and management software. After early tests in Australia, the system was selected to support Sydney’s World Youth Day in 2008, including a visit by Pope Benedict XVI, according to David Young, a spokesman for DigiSensory USA.
Now Avista systems are popping up in other parts of the world, from the United Kingdom to Louis Vuitton stores in China. Although the EOPD was the first U.S. agency to adopt the system, Young said a New Jersey housing authority is installing a 65-camera system and other agencies have been researching the product.
The UK Olympic organizing committee is also evaluating the camera system with London’s Metropolitan Police’s Victoria Station because it’s interested in using the technology to support the 2012 games. While no other law enforcement agency has yet installed the East Orange model, the system has piqued the U.S. Justice Department’s interest, Young added. From crowd control to public safety, the company sees tremendous public-sector potential in the system’s possibilities.
“The technology has wide application in law enforcement, school safety, government institutional safety, crowd management, transportation departments, prisons and more,” Young said. “With bullying being one of the biggest problems occurring in American schools today, imagine if principals would be automatically notified of a potential incident moments after students began to loiter.”
With sensors connected to surveillance cameras, the system is designed to identify specific behavior patterns that are linked to probable crimes: strange loitering, a group chasing an individual or a car creeping around the corner near a pedestrian. Crimes are recorded and stored in a database, so the system is always getting smarter.
But how can these sensors know the difference between a crime and a random, innocent act or gesture? That, Kenney said, is the age-old paradox.
“The problem is creating software that is able to examine human interactions and make determinations whether crime is occurring or not,” he said. “What if someone is walking up to anyone at the ATM and the system kicks on and says a robbery might be taking place, but there is no crime? It could wind up being a tremendous drain on resources.”
According to Cordero, avoiding this problem starts by ensuring that operators monitoring the video study the potential victim’s behavior.
“Someone who is in fear that they may be robbed or assaulted, they’re going to react a certain way,” Cordero said. “This is something that we need to look at to minimize false positives by not trying to re-create all variations of a crime, but looking at victims of the crime and their reactions and working our way back.”
You can’t have a predictive tool without running into the usual legal and ethical questions. For example: “Can you generate probable cause?” asked Peter Scharf, a public health professor with a specialty in criminology at Tulane University in New Orleans.
Scharf co-authored a study on gunshot detection systems used in Virginia. No matter how good forward-thinking technology may look, its success ultimately comes from credibility, Scharf said.
“You don’t want to kill innovation, but you want to structure it so it has scientific, ethical and legal control over technology,” he said. “Before you go screaming ‘This is it!’ you have to check it to make sure you’re right.”
With the EOPD leading the charge for law enforcement, the department has little research to support the system’s effectiveness. Cordero knows this, and even said the department sees itself as “the lab for what works” in the police force — trial and error comes with the territory.
As with the smart cameras, the EOPD realized soon after deployment that the caution alerts only gave patrol units a small response window, which Scharf considers a huge hurdle. “It doesn’t matter how quick the information comes. How quick can you get there?” he said. “Unless cars go faster, you’re not going to catch anybody.”
With the early tests, Cordero said, the EOPD ran into this problem when a sensor picked up a potential car burglary. When the patrol units arrived on scene, the suspect was gone. They eventually found him, but Cordero wanted to find a way to minimize delays. “If you can’t intervene,” he said, “then this is merely crime-recording technology and not helpful in preventing crime.”
In March, the EOPD launched an alarm-based automated dispatch system to work in concert with the smart cameras and cut down dispatch time. “From detection to validation, we have a car rolling within a second and a half,” he said. “Criminals see that when one of these systems goes off, the police are all over the place in a matter of seconds. That’s really where we drive crime reduction.”
You may use or reference this story with attribution and a link to