Excerpted from Vision Systems Design, December 2007

Vision systems tackle fingerprint analysis

Conventional fingerprint systems process inked fingerprint cards that have been manually scanned or fingerprints that are electronically digitized with flatbed scanners. To date, hundreds of millions of such prints have been collected and enrolled into systems such as the US Federal Bureau of Investigation (FBI) Integrated Automated Fingerprint Identification System and the US Department of Homeland Security (Washington, DC, USA; www.dhs.gov) Automated Biometric Identification System.

To digitize all ten prints from the digits of the hands, today’s capture technology requires approximately three minutes. In 2004, the Department of Homeland Security instructed the FBI to expand its fingerprint database, mandating the use of faster scanning times. That same year, the National Institute of Justice (NIJ; Washington, DC, USA; www.ojp.usdoj.gov/nij) solicited applications for a fast fingerprint-capture program to develop more efficient methods for collecting fingerprints.

“The NIJ has two basic requirements for fast fingerprint-capture systems,” says Laurence Hassebrook, associate professor of electrical and computer engineering at the University of Kentucky (Lexington, KY, USA; www.uky.edu). “First, the system must scan all the digits of one hand in less than 10 s, and second, it must generate an image of a rolled-equivalent scan—finger nail to finger nail—at 500 pixels/in. or better without the help of a human operator.”

Four independent projects for live-scan replacement are included under the NIJ Fast Fingerprint Capture Program using sensor technologies that are considerably different from either ink or flatbed scanner based systems.

…Both Carnegie Mellon University (CMU; Pittsburgh, PA, USA; www.cmu.edu) and the University of Kentucky are developing camera-based systems that capture the 3-D shape of the hand and the friction ridges of the fingers and palms…

Rather than use spotlights to illuminate the hand, Hassebrook and his colleagues at the University of Kentucky rely on structured light. By projecting and capturing a series of striped patterns over an object, its shape can then be deduced by analyzing the way the stripes warp over the object’s surface when viewed at an angle by a camera. In Hassebrook’s system, the hand is scanned, a 3-D image generated in real-time and then converted to simulate a 2-D rolled fingerprint.

Developed in conjunction with FlashScan 3-D (Austin, TX, USA; www. fl ashscan3D.com), the initial prototype scanned a single finger. The subject places a finger over an opening in the systems enclosure. A digital light projector projects a series of striped patters onto the finger’s ridges. An array of three Camera Link cameras from Basler (Ahrensburg, Germany; www.baslerweb.com) then acquires a series of images that wrap around the finger. These images are transferred to a host PC using two Solios XCL video-capture cards from Matrox Imaging (Dorval, QC, Canada; www.matrox.com/imaging).

A second prototype uses a single 4Mpixel Camera Link camera with a single Matrox Helios capture card to acquire a scan region large enough to capture a human subject’s palm. The system can also acquire the prints of all four fingers simultaneously but does not achieve wraparound scanning of any finger. To achieve instantaneous acquisition, future prototypes will feature a single, continuously projected custom designed composite pattern instead of the projected striped patters and higher-resolution cameras for simultaneous acquisition of the entire hand with wraparound finger scanning. After the system is completed, it will be tested by the Kentucky State Police and then delivered to the NIJ for further evaluation.