China’s traffic police are using AI to tackle “Chinese-style jaywalking” at major urban intersections. Facial recognition cameras take a 15-second video and four snapshots of pedestrians crossing on a red light. Pictures are matched with photo IDs in the police database, and violators can have their headshots along with family name and partially obscured citizen ID and registration address displayed on large roadside screens.
Since deploying the system in April 2017, Shenzhen traffic police have caught over 13,930 jaywalking pedestrians and non-motorized vehicles. Jaywalkers are fined up to CNY¥20 (US$3) and are subject to traffic rule refresher courses or community service at road intersections.
“Since the new technology has been adopted, jaywalking cases have been reduced from 200 to 20 each day at the major intersection of Jing’Shi and Shun’Geng roads. Fewer people are crossing roads during red lights,” said Li Yong, a Traffic Police Officer in the eastern city of Ji’Nan.
Chinese-style jaywalking is a social nuisance. According to Ji’Nan city statistics, barging pedestrians and non-motorized vehicles account for 16 and 33 percent of traffic accidents per year respectively. As ever-broadening Chinese roads can now have 10 lanes at urban intersections, ignoring traffic signals is extremely dangerous.
Traffic authorities also plan to build a social credit system wherein jaywalkers will start receiving text messages or Weibo notifications. Traffic police will record the number of violations, and a certain threshold will affect the offender’s social scores, which may limit their ability to borrow from banks.
Some may ask whether this AI is saving lives or infringing on personal privacy? Li Yi, a research fellow at the Shanghai Academy of Social Sciences, says the public display of offenders’ photos and partial personal information may prove to be effective in reducing pedestrian accidents and injuries, “however, we always need to find a balance between law enforcement and privacy protection.”
Journalist: Meghan Han | Editor: Michael Sarazen