University of Minnesota Driven to Discover
Universities MUniversities Wordmark
Center for Transportation Studies Heading

Webinar Available

Camera networks for security and traffic applications

Nikolaos Papanikolopoulos, Professor, Department of Computer Science and Engineering, University of Minnesota-Twin Cities

November 4, 2010

Security personnel traditionally monitor thousands of cameras at one time, often viewing only a single frame from each network camera every few minutes. Such a system provides a large amount of data that is difficult to manage and is often ineffective in identifying potential threats.

At the November 4 ITS Institute Advanced Transportation Technologies seminar, computer science and engineering professor Nikolaos Papanikolopoulos discussed the applications of a camera network system that has been under development by University of Minnesota researchers since 2004. The system, developed with seed funding from the ITS Institute, allows a single user to successfully monitor an entire camera network.

Work on the system began in 2004, when Papanikolopoulos and his research team completed a project using a camera system to detect drug deals at bus stops. Further advancement took place when the Department of Homeland Security asked the researchers to install a camera network at the 30th Street Station in Philadelphia, where all cameras and sensors throughout the transit facility were routed to a single individual for monitoring. A similar system will soon be installed at the light-rail station of the Minneapolis-St. Paul International Airport to observe human activities and perform threat evaluation, Papanikolopoulos said.

The camera network system is based on the Hyperion framework, which uses human and crowd activity monitoring, automatic camera placement, camera-to-camera tracking, semi-autonomous calibration, and video forensics analysis to evaluate all incoming video feeds and pass on critical information to a human operator, Papanikolopoulos explained.

For example, the system is capable of detecting a thrown or abandoned object anywhere in the network and informing the user of its presence. The user may then examine the data and choose to further investigate or ignore the incident. Additional system capabilities include trajectory identification, detection of camera tampering, and crowd monitoring. A search feature also allows the user to find all frames containing a particular individual, eliminating the need to examine footage from multiple cameras frame by frame.

By calculating the most efficient locations to place cameras, the system also allows a network to provide increased accuracy using fewer cameras, Papanikolopoulos said. This reduces the amount of duplicate or unnecessary footage and also keeps maintenance costs down.

The system’s wide-ranging detection capabilities make it an effective tool for a variety of transportation applications, according to Papanikolopoulos. The system can be used at traffic intersections to evaluate the trajectories of pedestrians, count the number of bicycle crossings, or collect other data. Research is also under way on using the system to identify empty parking spots at truck stops and direct drivers to locations with space available.

Papanikolopoulos also explained that possible system applications extend far beyond security and transportation issues. For example, the system will be used in the future to monitor the mental health of children in daycare, with a focus on identifying precursors to various disorders. System technology could also be employed in radiation therapy, Papanikolopoulos said, where it could help machines identify patient movement to avoid damaging healthy tissues.

“This is an area of vast research and deployment opportunities,” Papanikolopoulos said. “The implications are tremendous.”