Industry gets shot at face recognition

Officials at the federally funded Office of Law Enforcement Technology Commercialization (OLETC) are seeking a private company to commercialize a 3-D facial recognition technology that might someday be coupled with airport security cameras to catch terrorists.

The technology, called the Integrated Law Enforcement Face-Identification System (ILEFIS), creates 3-D images of human heads by using existing 2-D images, such as a police mug shot and a profile shot, potentially making it easier for law enforcement officials to recognize a match.

ILEFIS requires a lot of development, as it exists only in a beta version, said Arsev Eraslan, chief scientist of the ILEFIS project.

Eraslan said OLETC officials are searching for a private firm that could license ILEFIS, improve upon it and market it, Eraslan said. "We want to get it out of the door as soon as we can," he said. But he also noted that the firm selected to commercialize ILEFIS must focus on developing applications for the product to meet law enforcement needs.

Software on the market today already can search through thousands of those 2-D images, like driver's license photographs, to help law enforcement officers locate a criminal suspect. The software compares one frontal or profile image with another frontal or profile image to find the wanted person.

But searching a database of images taken at various angles - such as images taken from airport security cameras - proves a challenge when law enforcement officers compare the images with photographs taken from frontal and profile views.

"If that guy's a bad guy, and he's going through video surveillance, you're not going to catch him very easily," Eraslan said. Eraslan and other researchers said they believe the ILEFIS should ease the task of matching faces with images that have been taken at various angles.

Having 3-D images created from police mug shots should mean that law enforcement officers or investigators searching through a database of photographs or video images taken at, say, a 45-degree angle, would be able to generate an image of a wanted person's face at the same angle. Software would compare the images and show investigators potential matches.

ILEFIS works by focusing on 64 facial features - a chin, a nose, a lower lip or a left eye, for example. Eraslan said researchers have identified as many as 256 unique shapes for each of the features. Researchers process ILEFIS images of faces using a set of numbers. "No. 21 nose, No. 32 eye socket, No. 34 lip," Eraslan said, describing a possible combination for a human face. "So what we get is a head 'code,' " he said. Having numbers assigned to each part of the face - instead of conducting elaborate geometric processing of facial images - will make for quicker processing and searching of images, according to Eraslan.

OLETC will demonstrate the product for the Justice Department's National Institute of Justice next month. OLETC is a program of the National Institute of Justice and is located at the National Technology Transfer Center at Wheeling Jesuit University in Wheeling, W.Va.

Keith Angell, president of facial recognition software company Miros Inc., said he would have to see ILEFIS demonstrated before deciding whether his company would want to license it. He said attempts to combine 3-D technology with facial recognition have not been successful. "I would think that they would have a hard time making that functional quickly," he said.

Frances Zelazny, marketing manager at facial-recognition software company Visionics Corp., said current facial recognition technology has difficulty matching faces beyond a 35 degree angle. "Beyond that, you lose accuracy."