A recent study conducted by researchers at MSU was able to evaluate the latest in automatic facial recognition technology — programs that quickly can attach a name to a face by searching a large database of face images and finding the closest match.
Using three different facial recognition system tests, MSU computer science and engineering professor Anil Jain and research scientist Josh Klontz were able to use one of the Boston Marathon bombing suspects to demonstrate this technology in law enforcement.
“I was very pessimistic (about) finding correct matches (at first), given how low quality the images people were using,” Klontz said. “The fact that we got one right match, and it was the first image was promising in terms of actually being able to solve this kind of problem in the future.”
Within the Pattern Recognition and Image Processing Laboratory, Jain said Klontz and himself were able to use law enforcement video from the bombing to provide a match of suspect Dzokhar Tsarnaev.
“It can be easy or difficult depending on the quality (of the footage) and (in) what conditions they are captured,” Jain said. “The simple case is a mug shot (or) driver’s license and these are captured in a controlled condition (when) your expression is neutral and the background is constant.”
When pictures are being taken in a controlled environment, the degree of accuracy is 98 to 99 percent, according to Xiaoming Liu, an assistant professor in the Department of Computer Science and Engineering.
For the past few decades, the facial recognition programs have been designed to be used in controlled environments, such as when people are facing the camera, they’re not smiling and the lighting is evenly distributed. Liu said the concern becomes apparent in situations such as the Boston Marathon bombing, when the suspect can be hard to see.
“When people are captured in these situations, they can wear sunglasses, (they) can have different kinds of poses (and) the lighting can be bad,” Liu said. “In these conditions, the chances of finding them (are) less.”
In these unconstrained environments, the accuracy percent is closer to 80 percent, but he hopes to see that percent increase in the next few years, Liu said. This will enable local law enforcement, such as the MSU Police Department to become more involved with these programs as well.
“This kind of technology can be developed furthered and law enforcement might able to see an advancement and be able to review this kind of information,” Klontz said. “What we were hoping to do is to draw attention to the possibility of this kind of technology and how it could be available in the near horizon.”
Support student media!
Please consider donating to The State News and help fund the future of journalism.
Discussion
Share and discuss “MSU researchers use facial recognition software to identify Boston Marathon bombing suspect” on social media.
