PHILADELPHIA (CBS) — The brains behind surveillance and in-car cameras are getting a lot smarter: they’re designed to help cops track criminals and keep police cruisers on the cutting edge, right when officers need it most.
Artificial Intelligence was one of the technologies generating plenty of buzz at the International Association of Chiefs of Police convention in Philadelphia.
“This is a trained and learning solution that allows us to identify vehicles on the fly, in real-time,” says Mark Griffin of Coban Technologies.
He’s talking about the Focus H1: AI baked into the dash-mounted camera makes it more of a partner to police.
“For example, say it’s an amber alert, and you’re looking for a white Ford pickup truck,” he says. “Or you may have part of the license plate number.”
(Ian:) “This is a patrol car, driving through a parking lot, and the camera screen shows it immediately recognizing — hey, that’s a Ford pickup truck, that’s a Toyota Corolla, and it’s grabbing the license plates, the colors, and all that stuff?”
(Griffin:) “That’s correct. All that metadata. And what you get from that is the capability of getting that BOLO identified very quickly.”
Criminals should “Be On the Look Out” for this in-car camera tech early next year. It also can identify faces, weapons, certain behaviors, or be trained to flag pretty much anything else cops can come up with.
Here’s another hypothetical: an attack inside a train station. Investigators pull the surveillance video, and flag suspicious characters. But how do they track the suspect across a sea of people?
“You can pick a person and say, ‘show me where this person appears across large amounts of other cameras,'” says Tom Graham with the video intelligence company Seequestor. “You can search for that person much faster and save yourself the time.”
Faces in CCTV footage often can be too blurry to be useful, so the company’s software algorithm analyzes full-body images.
“What were they up to scouting the place for maybe a week beforehand?” Graham says. “So they’ll extract an entire week of video from the train station and have to find them within it.”
That’s a needle in a 21,000-hour haystack: police in Australia used Seequestor to zip through that much video in a fraction of the time.
“They were able to walk away with what they needed in four days, which was an unthinkable achievement,” Graham notes.