Store Surveillance Keeps an Eye on Your Hair and Clothes
Retailers can now use software to track one person across multiple surveillance cameras, possibly helping them improve store layouts.
A startup is trying to make it easier for store owners to figure out what people tend to do in their shops. Its software examines surveillance footage and tracks customers as they move around based on what they’re wearing or the length of their hair.
Boston-based Netra takes surveillance footage from stores’ Internet-connected video cameras, breaks it down into different components, and extracts characteristics—such as long brown hair, a plaid shirt, blue jeans, and so on—about any person spotted on the video, and stores it in an index. As that person moves around the store from, say, the produce aisle to the cereal aisle, Netra’s software determines how likely it is that the person spotted in one place is the same person later spotted in another.
The technology, which grew out of cofounder and chief technology officer Shashi Kant’s graduate work at MIT, is still in its early days; the company says it’s just starting pilots with some retailers, including one fast-food chain. But Kant says a study the company conducted using the technology indicated it was around 80 percent accurate, depending on lighting and camera contrast. This could make it a useful tool for retailers trying to improve things like the store’s layout or product selection.
Netra’s software picks out people’s characteristics from surveillance video so it can track them across a store, from one camera to another.
Figuring out what’s happening within surveillance videos without the help of a human can be tricky. Netra isn’t the only startup trying to sort out customer traffic and activity—other startups, like Prism Skylabs (see “Surveillance Video Becomes a Tool for Studying Customers”) and Density (see “A Sensor for Logging People Traffic at the Gym or Café”), are trying to do so, too. But CEO Richard Lee thinks Netra can capture more accurate information by extracting color, texture, shape, and contour from objects that store cameras see, and by tracking these things from one camera to another.
The company plans to start by tracking store traffic, the time people spend in stores, and their activities, Lee says. In addition to helping retailers figure out traffic patterns, he says, Netra’s technology could give them a better sense of how productive their salespeople are.
Netra—like others who have used surveillance technologies—may find that shoppers are wary of being tracked, even though Lee says it’s not using facial recognition and the information is anonymized. Brian Mennecke, an associate professor at Iowa State University whose research includes surveillance and privacy, points out that while cameras are already in stores, we tend to think they’re being used to keep an eye on anyone who might be stealing, not our hair or clothing.
However, he says, if stores can find a way to present the technology to customers as something that will help them find a product, they might be okay with it.
“It can always be presented in a way where people perceive there’s a benefit,” he says.
Hear more about image recognition at EmTech MIT 2017.Register now