Investigators in Boston this week quickly began sifting through more than two thousand videos and still images of the marathon route looking for potential suspects in bombing, with crowd sourcing becoming the newest tool in the arsenal of law enforcement agencies.
But researchers in upstate New York are working on a new technology that could make it easier to analyze the enormous volumes of visual content available from devices like smartphones and surveillance cameras.
“We want to be able to provide an ability for the computer systems to readily summarize the large amounts of video that you see out there,” says Dr. Eli Saber from the Rochester Institute of Technology.
Saber says the goal is to develop complex computer algorithms that will enable a system to recognize aspects of a scene that are potentially incongruous or out of place.
This could include recognizing an unattended bag at an airport as suspicious and sending an alert to security.
One of Saber’s colleagues, RIT’s David Messinger, says this could save humans from having to sift through hours of video footage frame by frame. He says the human brain is great at understanding an image, but there are some areas in which computers are more efficient.
“The computer is also very good at other things, processing large amounts of data, extracting features from that data – not necessarily understanding what those features mean – but being able to pull things out of the data that you can then provide to the human brain system which is very good at understanding the meaning of things,” said Messinger.
The technology has applications ranging from security to disaster response and fast-tracking medical diagnoses.
Saber and Messinger have been awarded over $1 million in grants from the Department of Defense to continue advancing this technology.