Videos taken at the crime scene could hold important clues. But it's a different kind of fast-forward investigators can now use to review all the footage. CNET's Kara Tsuboi goes face-to-face with facial recognition and surveillance.
After the Boston Marathon bombings, police in the city made a plea for people with cell phone video and pictures to turn over their footage, adding to the hours of surveillance video from nearby businesses. But what would normally take investigators hundreds of hours to review can now take minutes or even seconds, thanks to technology like facial recognition. The software, which can help pick a person out of crowd, looks for differentiating features -- from the shape of a mouth to the ridge on a nose to the distance between a pair of eyes.
3VR in San Francisco has developed software that extracts information from video and then makes it searchable for its clients, which include retailers, banks, security firms, and law enforcement.
The video stream can come from surveillance cameras and smartphones. "We will identify each person and extract the facial biometrics of each person in the field of view, and we'll save a snapshot of that person," explains 3VR CEO Al Shipp.
The biometric information is turned into metadata that's searchable using different filters. So investigators can focus only on certain people while quickly and easily eliminating those that don't fit the suspect's description. However, facial recognition isn't just for crime fighting. Some retailers and marketers are exploring its use to learn more about customer behavior and develop personalized marketing.
The expansion of the technology is raising some privacy concerns, according to CNET Senior Writer Declan McCullagh. "What can happen is you can have this idea of cradle to grave surveillance when you're out in public. Storage is very cheap. It's possible for surveillance cameras and facial recognition technology to keep a record of where you are and where you go wherever you're out and in range of cameras."
CNET took a closer look at facial recognition: