Researchers at IBM Corp. are trying to address privacy concerns about video surveillance systems, part of a broader effort by IBM to build a new business in the fast-growing surveillance market.
Concerns about security in cities, airports and other public places are causing a proliferation of video surveillance systems, but the increase has heightened concerns about privacy among regulators and the general public.
IBM hopes to alleviate the concerns with technology that can pick out faces in a video frame and automatically blur them, so that people’s images — and therefore their movements — are not recorded, said Joachim Stark, director of digital video surveillance with IBM’s global services group.
An obvious hurdle is identifying the potential suspects from innocent bystanders. Investigators often review closed-circuit video footage after a crime is committed, and blurring faces would defeat much of the point of doing surveillance.
One solution is to find ways to identify suspects automatically so that only their faces are left unblurred. Video analytics software can already trigger an alert when a person leaves an object of a certain size on a station platform, for example, and walks off. After spotting such a behavior, a surveillance system could “rewind” the action in Tivo-like fashion and unblur a suspect’s face from the moment the person enters the frame, Stark said.
Another option is to blur all the faces when the video is recorded, but allow investigators with the right access permissions to unmask them at a later date.
None of the solutions are perfect, and Stark said it’s likely to be a few years before the blurring technology, being developed at IBM’s T.J. Watson Research Center in New York, is ready for commercial use. Another hurdle is being able to identify and blur faces in real time. A prototype was demonstrated at the Cebit trade show in Hanover, Germany.
In the meantime IBM has an array of software and services to pursue its video surveillance push, which began late last year and is aimed at the retail, banking and public sectors. It calls the products its Smart Surveillance System. It’s a newcomer to the market, and will compete with established players like Vidient Inc., ObjectVideo Inc., and others.
The video surveillance market is growing at around 15 percent annually, Stark said. IBM hopes to distinguish itself with its database and middleware technologies, which can help store and analyze the vast quantities of video data.
Surveillance technologies have already come a long way. IBM’s analytics software records metadata, or information about the data in a video, such as colors and the size of objects in a frame. If a witness reports seeing some on a red sweater acting suspiciously, investigators can search for “red” in the surveillance software and pull up the relevant images.
Such systems can generate vast amounts of data, however, and IBM is looking at compression technologies to reduce the volume.