Harry Potter fans, take heed. New surveillance software that can track the whereabouts of over a dozen people at the same time is providing researchers with their own Marauder's Map, allowing them to monitor the comings and goings of subjects in indoor settings nearly as complex as Hogwart's.
Researchers at Carnegie Mellon University in Pittsburgh have been perfecting their multi-camera tracking system- which they use to monitor nursing home residents- since 2005. The project was originally developed to keep track of elderly residents' health by monitoring their activity levels and behaviors.
But Alexander Hauptmann, one of the project's researchers, believes that the system could also be useful for identifying suspected criminals or terrorists, like the perpetrators of April's Boston Marathon bombings.
Much of the video analysis necessary for tracking such people is done manually, which is both time-consuming and imprecise. Automating tracking techniques would make them more useful for security personnel in airports and other public facilities, Hauptmann said.
And if used in conjunction with existing algorithms for determining who on a surveillance camera feed is a potential threat or even a wanted criminal, the CMU algorithm could go a long way toward improving public security systems.
CMU's tracking system uses a network of video cameras and carefully calculated algorithms that take into account multiple cues from the camera feed, including apparel color, person detection, trajectory and facial recognition. [See also: Military Wants to Detect Terrorist Body Language ]
Unlike other researchers studying multi-camera, multi-object tracking, the CMU team chose to test their tracking techniques on real subjects instead of in a virtual lab, which made their research more difficult but also helped them develop a product well adapted to the real world.
Their tracking system works even when faced with common obstacles like long hallways, doorways, people mingling in the hallways, variations in lighting and too few cameras to provide comprehensive, overlapping views.
The CMU system can locate individuals within one meter of their actual positions with 88 percent accuracy, a feat that Shoou-I Yu, a researcher with the study, thinks is possible largely because of facial recognition technology. Removing facial recognition from the algorithm's data dropped the system's accuracy to just 58 percent.
Researchers say their future plans for the tracking system include adding privacy measures- such as recording a person's outline instead of their whole body- so the system can be used less intrusively inside care facilities like nursing homes.
They also plan on adding the use of depth cameras, like Microsoft's Kinect, to more accurately determine subjects' location.