Video analytics based on clothing, general appearance more effective than facial recognition, says NICE Systems
What is in this article?
Video analytics based on clothing, general appearance more effective than facial recognition, says NICE Systems
When a woman saw a man steal an iPad from a public transit passenger, the main thing she noticed was his orange shirt.
That detail and the time of the incident were enough for the authorities. This undisclosed U.S. public transit system is one of the proof-of-concept (POC) deployments of NICE Systems’ recently announced Suspect Search video analytics system.
Using the system, the transit officials were able to scan surveillance footage for people wearing orange shirts, identify the culprit and then track his movements on a map.
“It’s not a real exciting use case—you know, iPads—but it proved the technology,” Chris Wooten, executive vice president of surveillance solutions for NICE, said during an interview with IWCE’s Urgent Communications.
These POC deployments, which have been operating for about four months, have already resulted in purchase orders, but Wooten declined to say how many have been sold at this point. Suspect Search, which was showcased at the ASIS conference in Atlanta this month, will not be generally available until the first of the year, he said.
“There’s been a promise of analytics in this industry that’s never really come to pass, so we said this time—before we get too crazy about it and do a big splash—‘Let’s go out and put it in some customer locations and make sure it works,’” Wooten said.
“And the testimony has been, after we’ve done these proof of concepts, we get purchase orders. What we do is we go set up a system at their location, connect it to their cameras, and they tell us how they want to run the POC so it’s not like we’re doing smoke and mirrors.”
NICE Systems currently has POCs deployed with a major U.S. airport, a military hospital that has high-profile visitors on campus, multiple mass-transit organizations and Glascow, Scotland, where city officials had a particular interest in using the system to locate lost children quickly.
The price for the solution depends on the number of video channels and the services required for installation, including software installation, a review of cameras for angles and image quality, and training, according to a company spokesperson.
Suspect Search is not a facial-recognition application, stressed Doron Girmonsky, NICE System’s head of technology and innovation. Surveillance cameras are not ideal for facial recognition, because a clean, full-frame image of a person’s face is needed, and—in most cases—surveillance cameras offer frames packed with multiple people. Also, there are already advanced facial-recognition solutions on the market, he said.
Wooten said he also believes searching for clothing can be more effective than facial-recognition technology.
“It’s much harder to remember what their face looks like as opposed to what they were wearing, and it actually gives us a higher success rate by understanding what they were wearing and roughly what they look like,” Wooten said.
Suspect Search scans hours of footage, indexes the human images and compiles them in a database. Each image is given a unique signature that identifies the person in the image and the time the image was captured.
When conducting a search, a user can leverage an existing image or create an avatar of the individual, piecing the cartoon image together based on what is known about the person’s appearance. The user also can narrow the footage and manually search, until an image is found that can be used to find more. After finding a relevant image, the user can watch the accompanying video for context.