Startup Scans In-Store Surveillance Footage to Analyze Shoppers' Preferences

Surveillance Cameras
Jonathan McIntosh via Flickr

Millions of security cameras capture constant video at businesses and retail locations throughout the U.S., but for the most part their footage is only useful if someone shoplifts and cops need to check it out. But there's a wealth of data buried in that video, from customer density to crowd shopping preferences. A new startup can analyze surveillance video to help business owners see what their customers do, in the way websites can easily track online shoppers' browsing habits.

Prism Skylabs, based in San Francisco, installs software on computers linked to existing security cameras. The program uses computational photography — sort of like the Lytro light field camera — to produce images with higher resolutions than the original grainy CCTV video, and then edits out people's identities for privacy's sake. Humans appear as ghostly figures or are edited away completely, leaving colorful discs in their place that depict a crowd's size and density.

The goal is to monitor traffic, to shed light on how people move around a store and even to gauge the public's interest in certain products, by studying where people gravitate and linger in a business. "It's like Google Analytics for the real world," Steve Russell, cofounder and CEO of Prism, told Technology Review.

It could also be used to help the public, too, by showing crowd sizes in real time at a gym or a restaurant, according to Tech Review. Video can work better than photos to give a sense of place — in some ways it's surprising no one else has tried this before.