Osaka train station set for large face-recognition study

Japan’s Osaka Station could become another focal point in the global battle over personal privacy protection as a Japanese research center prepares for a long-term face-recognition study there.

The independent research group National Institute of Information and Communications Technology (NICT) plans to begin the experiment in April to study crowd movements in order to better plan for emergency procedures during disasters.

The train station is western Japan’s busiest, with an average of 413,000 passengers boarding trains there every day. Over a million people use it and neighboring Umeda Station daily.

NICT will deploy cameras in Osaka Station and the adjacent Osaka Station City, a multipurpose complex, that can track faces as they move around the premises. The cameras will be separate from any security cameras that are already installed by operator West Japan Railway (JR West), a spokesman for the railway said.

"The purpose of the study is to determine whether or not sensor data on crowd movements can be used to validate the safety measures of emergency exits for when a disaster strikes,” a NICT spokesperson said.

JR West referred all questions about the protection of passenger privacy to NICT, but the research institute said the experiment was still being prepared.

"At this time, we are considering the technology to be used to obtain statistics on crowd flows,” the NICT spokesman said. “Depending on the technique used, the data that can be obtained on pedestrian flows will be different. So, it’s difficult to say how many people could be subject (to the experiment).”

NICT would not elaborate on technical aspects of the study, but it said previously that it is slated to run for two years and will involve about 90 cameras and 50 servers.

It said the facial-recognition system can track dozens of points on a face. It emphasized that the data cannot be used to identify people and it will abide by Japan’s Personal Information Protection Law when handling it.

The NICT spokesperson said the institute is not aware of any similar large-scale study.

A report in the Yomiuri Shimbun newspaper said computers linked to the cameras will run face-recognition algorithms and assign IDs to faces, which will be tracked for a week.

The dozens of cameras would be able to track the movements of people the algorithms recognize, and record whether they go to a coffee shop or through a ticket gate, for instance.

The technology under consideration for the study can identify faces with an accuracy of 99.99 percent, according to the report.