Facial Recognition Algorithm
Smartphones and other mobile devices equipped with facial recognition software can already identify individuals under ideal conditions, such as having a good-quality photo of the person in a database with which to compare in real time. However, such optimal image conditions are pretty rate. In addition, people’s faces change over time, and donning a pair of sunglasses or growing a beard can prevent the technology from making an accurate match.
And although videos should provide a better chance for forensic scientists to identify a suspect since they offer a series of images, this too doesn’t always happen, as was the case of the Boston Marathon bomber: In a test of three facial recognition systems, only one identified Dzhokhar Tsarnaev, and none of them recognized Tamerlan Tsarnaev, who wore sunglasses.
However, a company called Animetrics may have the answer to these problems. Animetrics has company has developed software that converts 2-D images into “simulated 3-D models of a person’s face” in about a second, and the software’s users can alter a suspect’s attitude or position. The resulting “headshot image” can be analyzed by all facial recognition algorithms. On a sophisticated laptop, the headshot can be matched against as many as one million faces. For smartphones, the algorithms must be scaled down, which makes them less effective. Experts are confident that the limitations of smartphones can be offset in the future by using the cloud to compute the algorithms. Then, the technology will fit into the palm of a police officer’s hand, allowing almost instantaneous identification of suspects.