SECURITY/Facial ID system scans visitors at popular beach
To augment the video monitoring of the city’s oceanfront, the Virginia Beach, Va., Police Department has installed a facial identification system. The technology examines facial images captured by three video cameras near the beach and compares them with approximately 650 pictures of people that the Police Department is trying to locate, such as those wanted on felony charges and missing children.
About a decade ago, the Virginia Beach Police Department installed 10 video surveillance cameras throughout the area, which brims with both residents and vacationers during the summer. For instance, in 2001, approximately three million people visited the city, which has a population of 425,000. With tourism playing a vital role in Virginia Beach’s economy, the city wanted tourists to feel its oceanfront was safe.
Using a closed-circuit television network, the cameras feed live video images into a control center located near the oceanfront. Videotapes of the feeds are kept for a period of time and then recycled.
No incident triggered the purchase of the facial identification technology, says Greg Mullen, deputy chief of investigations with the Police Department. Rather, the city installed the video monitoring system in 1993 with the expectation of expanding it as new technologies became available, he says.
In 1999, Virginia Beach Police officials saw a demonstration of facial recognition technology at a conference. “We started doing some research on it, and we were able to learn that the system would be very compatible with the system that we already had in place,” Mullen says. “At that point, we started moving forward with the upgrade.”
Virginia Beach purchased FaceIt ARGUS, facial recognition software manufactured by Minnetonka, Minn.-based Identix, from a partnership of two vendors, locally based Conseps and Miami-based ATC International Holdings. The two vendors also provided three video cameras and fiber optic cables. Installation began in July 2002, and the system was fully operational two months later. In all, the upgrade cost $200,000. A state grant paid for $150,000 of the costs; Virginia Beach paid for the rest.
Located on a computer in the control center, the software uses a mathematical formula to examine the spatial relationships between points on a person’s face and compares the results with images stored in the database. If the system determines that a match has been found, an alarm sounds, and officers print the two images. At that point, officers are dispatched to investigate further. The system discards all unmatched video images immediately. During testing that the Virginia Beach Police Department performed last summer during daylight and dusk hours, the system produced a correct alarm rate of 87 percent.
As Mullen explains, there are limits to the system’s ability to examine facial images. “The system’s looking for a certain number of points so that it can recognize a face as a face,” he says. “Obviously, if your head is turned and it can’t see the front of your face, then it’s not going to pick up the image. If you have a hat on that’s covering your face, then it’s not going to pick up the image.”
The Police Department has not made any arrests through the use of the system, and Mullen admits that the chances of doing so are not great right now. “You’re only going to make an arrest if a person who’s in that database walks through one of those camera fields. At this time, we only have three cameras that are feeding images into this database,” he says. “But, we see it as another tool that we can use to try to identify people who are wanted and provide some safety to those folks who are visiting the area.” Mullen adds that regular testing of the system indicates that it is functioning well.
A future expansion of the identification system to include more cameras is a possibility, Mullen says. The Police Department will spend the next year further evaluating and testing the system. “At the end of that time, we will make a decision if we want to expand it and determine if there are other mechanisms that we need to look at to improve it,” he says.