Gotcha! Recognizing Faces, AutomaticallyBy Sean Gallagher | Posted 2004-04-12 Email Print
WEBINAR: On-demand webcast
Next-Generation Applications Require the Power and Performance of Next-Generation Workstations REGISTER >
Variations in lighting can be enough to thwart a viable match.
Casinos have had some success using software and video cameras to recognize card and slot-machine cheats. Picking the right equipment and putting it where it will work best is often the difference between catching thieves-or terrorists-or letting them walk through the gates.
PROBLEM: Systems often have a hard time scanning faces in a crowded, busy place.
RESOLUTION: Line people up. At sporting events, cameras can be placed at ticket turnstiles. At airports, the security lines are a natural location. Operators searching for a particular individual also can enhance the systems' effectiveness by limiting the parameters of a search-by gender, race or other characteristics that may be stored in a facial-recognition database.
PROBLEM: It's hard to find an exact match.
RESOLUTION: Don't use a system that is geared to produce an exact match. Look for a system that presents a list of possible matches ranked from highest to lowest probability. These systems require the users-casino surveillance professionals or Transportation Security Administration agents-to compare the captured image against a virtual lineup of possible matches.
PROBLEM: Variations in lighting can make it hard to capture a usable image.
RESOLUTION: Use cameras that don't depend on ambient light. Some casinos have installed cameras that use near-infrared and low-power infrared lighting to illuminate surveillance areas. They work especially well when the images in the reference catalog also were gathered with infrared cameras-the system is then able to pick up and match things such as veins under the skin that aren't visible in normal light. But this isn't perfect. Infrared cameras only work at a range of about 20 feet-strong doses of infrared lasers could injure the eyes of those being watched.
PROBLEM: The database used for making matches might have photos of the wrong people.
RESOLUTION: Use multiple databases. The facial databases and outside information sources that casinos rely on aren't 100% reliable. Former professional gambler Roger Williams says that his friends got copies of their files in the database supplied by Griffin Investigations-a private firm that provides casinos with information on players-and found that some facts were wrong. Many casinos avoid the problem by cross-checking their files and photos against other sources, such as the Casino Visual Identification database from Biometrica, which contains a collection of known cheats, and by sending queries to other casinos.