3-D face recognition technology guards GSA

The Homeland Security Department has completed a successful pilot program that uses 3-D biometric facial recognition technology to control access to General Services Administration buildings.

The Federal Protective Service, a DHS agency that protects GSA buildings nationwide, is using equipment from A4Vision to identify and verify identities of people using one entrance to GSA regional headquarters in Auburn, Wash., said Ryan Zlockie, director of product management at A4Vision, today.

FPS recommends security practices and technology for federal agencies that lease space in GSA buildings, said Don Bowman, FPS’ acting regional director, in a statement.

“FPS evaluates electronic security products, which at times include biometric products, because some of our client government agencies require biometric identity verification. These agencies rely on our security assessments,” Bowman said. “Our evaluations of biometric security had been disappointing until we implemented A4Vision.”

FPS has used the A4Vision Vision Access 3-D Face Reader since October 2005 and authorized today’s announcement after a successful trial period, Zlockie said.

A4Vision’s 3-D facial recognition equipment shines a grid of invisible infrared light as far as six feet away from the machine, Zlockie said. When the subject walks into the grid, the light is distorted into the shape of the person’s face.

A video camera takes multiple frame pictures of the distorted grid and sends the video stream to a computer, which uses A4Vision software to analyze the images and build a representation of the person’s face, Zlockie said. The image is then compared with a control image stored in a central database, just like any other biometric.

A4Vision’s technology allows users to identify subjects in poor or no light and at an angle, Zlockie said. Its error rate is similar to fingerprints and better than 2-D facial recognition, he said.

The 3-D readers track parts of the face that don’t change, such as the curve of the forehead and the shape of eye sockets, Zlockie said. Most beards and eyeglasses don’t affect the machines’ accuracy, he said.

Users can view a 3-D mesh image, which looks like a face made out of window screens, or a smooth 3-D surface image, Zlockie said. If they want, users can overlay a 2-D color photo over the 3-D surface to get a full-color 3-D replica of the subject.

The technology allows quick throughput, taking the same amount of time as someone swiping an identity card through a card reader, Zlockie said.

The company uses commercial) technology with a proprietary algorithm, Zlockie said. It runs on Microsoft and Linux software.

Two-dimensional facial recognition technology creates a template built from analysis of the distance between a person’s eyes, ears and other facial features, Zlockie said. It also looks for distinguishing characteristics like scars.

The technology is limited to use in visible light and subjects must be viewed from certain angles for the technology to work, he said.