convialsa.org

NJBIZ, Feb 9, 2009 by Ruth, JoĂŁo-Pierre S

Safer alternatives in intelligent surveillance

TECHNOLOGY

PRINCETON – Surveillance cam- eras can spot suspicious activity as it happens, but after cease- less hours glued to the screen, security personnel watching them might overlook a potential threat. Some local researchers are working to give military defenders and security staff tools to be more aware of the action on screen as it happens.

Sarnoff Corp., Siemens and Princeton University are independently pursuing video systems that teach computers to automatically recognize action on screen; all are based locally or have operations in the Mercer County township.

Mike Piacentino, technical director of vision systems at Sarnoff, said the firm is producing surveillance technologyforthe Marine Corps. The Computer- Visio n Assisted Combat Capability or CVAC2, began development in the second quarter of 2008. It integrates information from 29 different sensors – such as video, sound, infrared and lasers – into one interactive display that feeds visual alerts to defenders. Piacentino said the device can be mounted on a standing tripod or equipped to a vehicle for mobile use; it was mounted onto a Humvee for the first time in January.

When a vehicle, person or other object moves on camera, the system tags the movement and displays an alert on screen, Piacentino said. CVAC2 also integrates a system called Boomerang, developed by BBNTechnologies in Cambridge, Mass., which detects weapons fire. “If there is a gunshot anywhere around you, we map that on to the visual display,” Piacentino said. “You don’t want war fighters sticking their heads out of vehicles.”

Last week, Sarnoff announced it was awarded a contract from the U.S. Marine Corps War Fighting Laboratory for a 12-month demonstration of CVAC2 in Quantico, Va.

CVAC2 was originally designed for perimeter surveillance, with multiple cameras networked together. “You can put one of these on each corner of a larger perimeter, and that allows you to stitch the video together,” he said.

The system is still a prototype, with a demonstration model due in the coming months, Piacentino said. A production model may be available later this year.

Piacentino said CVAC2 can “fingerprint” a person or vehicle that passes by a camera, storing information that will appear on screen if the object reappears. That can help better identify people or vehicles suspected of threatening activity. Piacentino said the addition of ground sensors for movement detection and other surveillance devices can make the CVAC2 video feed more comprehensive. “That’s the next generation of adding capabilities” to video, he said.

The system can be adjusted to ignore inconsequential movement, such as squirrels running by or tree branches swaying in the wind.

Sarnoff is not alone in developing advanced video technology. Fei-Fei Li, assistant professor in the computer science department, heads up the university’s Princeton Vision Lab, which researches technology to teach computers to recognize onscreen movement. Li and researchers are working on computer technology that includes Human Action Categorization, which learns to identify such movement as running, boxing and jumping.

“We are trying to understand motion types because it helps downstream to understand the types of activity going on in the environment,” Li said. The researchers at Princeton Vision Lab are focused on the basic research, Li said, using advanced algorithms to teach the system to understand visual inputs. She said research in this field could help car makers automate smart cars that would help avoid accidents by applying the brakes, or monitor swimming pools for people who are drowning