Hong Kong launches AI sensor trial to curb illegal parking meter occupancy
Hong Kong’s Transport Department has begun a six-month trial of AI-powered sensors at four metered parking spots to detect improper occupancy. The government is also considering integrating sensor data with payment records to create a proactive monitoring system, reducing reliance on public complain
The Hong Kong government is trialling new artificial-intelligence sensors at four metered parking locations to combat the illegal long-term or repeated occupation of spaces, the Transport Department announced on 28 October.
Since 2 July 2026, sensors have been installed at metered spots in Chai Wan, To Kwa Wan and Tsuen Wan. The sensors operate 24 hours a day, recording whether a space is occupied and using AI to analyse whether the object in the space is a vehicle, to assist enforcement authorities.
The trial locations were chosen based on factors including vehicle types, parking layout, and usage patterns. All four sites have average occupancy rates above 80 per cent and have previously received complaints about illegal occupancy, the department said.
Currently, the department relies on its meter operator to inspect each of Hong Kong’s more than 20,000 on-street metered spaces at least once every four days. Over the past three years (2023–2025), the operator referred an average of about 120,000 cases annually to the police, while the department received 111 related complaints — nine on Hong Kong Island, 64 in Kowloon and 38 in the New Territories.
Under the Road Traffic (Parking) Regulations, parking continuously for more than 24 hours in a meter space is an offence punishable by a fine of up to HK$2,000. Leaving objects such as pallets or goods to reserve a space can lead to a HK$25,000 fine or three months’ imprisonment under the Summary Offences Ordinance.
The new sensors do not collect vehicle registration numbers or facial images. Transport officials said they would study whether to combine sensor data with payment records and complaint logs to build a proactive monitoring system, rather than depending mainly on public reports. A review of the trial is expected by early 2027.