There is nothing worse than driving round and round a car park, unable to find a free space when the sign at the entrance tells you there should be 33.
Now Cambridge Consultants has developed a deep learning and machine vision system that can not only tell you how many spaces are left, but where in the car park you can find them. The AI system, dubbed Goldeneye, uses information from 12 security cameras to continuously monitor the availability of parking bays, in what the company dubs the world’s smartest car park.
Goldeneye was trained on footage of the car park in different lighting and weather conditions, day and night, teaching itself to recognise vehicles and both empty and occupied parking bays. Existing car park monitoring systems use sensors for each space, which can be expensive to install and maintain, according to Thomas Carmody, head of transport and infrastructure at Cambridge Consultants.
“Our approach was to take the existing camera feed coming from security cameras, and put it into a deep learning system to monitor individual bays, which is a much more cost-effective way to manage parking,” he said.
The system can be connected to the cloud, allowing users to choose a parking location, and reserve and pay for their slot online. The technology could also be used to monitor vehicle occupancy rates, footfall from vehicles and dwell time at retail locations.
“The use of deep learning and machine vision could be extended to other uses, so if you have this sort of system on a retail park, you could re-programme it to monitor pedestrian flow to and from the centre, for example,” said Carmody.
The technology could also be harnessed in smart cities, to monitor traffic flow or infrastructure upkeep, he said.