A new smart parking concept is offering an easier life for car park owners and urban drivers. Based on a video feed and a smart computer, the future of parking has arrived.
Every day, motorists waste hundreds of kilometres in their search for available parking spots, just in one car park. A new world-leading software system is now able to reduce this unnecessary traffic significantly by precisely identifying free parking spaces in real time, using an IP camera coupled to advanced Machine Vision, Deep Neural Networking and a clever back-end software suite.
One camera can cover up to 400 parking spaces, and there is no upper limit to the number of car parks and parking spaces that can be processed. In terms of distinguishing between cars and other moving objects, the system delivers unusually high accuracy, at about 98 %. At the same time, it has never been so inexpensive to carry out single space detection in large, open car parks.
This has been proven in tests conducted by COWI over the past 18 months. For the last 12 months, Jonas Hammershøj Olesen, Project Director at COWI, has been testing the system on behalf of Czech business collaborates, RCE Systems. According to him the results are radical:
"We've seen many different concepts over the years, but this is the best one - by far. Competitors are also using cameras, but their software is not smart enough to distinguish between, let´s say, clouds and cars. This makes their data unreliable. We've "trained" the software to know what a car looks like. This means that a cloud or a shadow will not mess up the data. We´re counting cars, and only cars. A classic system would cost ten times more than our system – and tests show that our system has an accuracy of about 98 percent," Jonas Hammershøj Olsen adds.
For the test, the COWI team set up two video cameras overlooking a car park in Lyngby, Denmark. The team then developed a variety of algorithms that enabled the Online Parking Occupancy software to distinguish between cars and clutter, using Deep Neural Networking otherwise known as Machine Learning.
In brief, a technology that, through hours of programming, enables a computer to get a high-level understanding from digital images or videos.
From the perspective of engineering, it´s possible to automate tasks that the human visual system can do. In this case, a video feed and a smart computer was sufficient to monitor and understand the dynamics of a car park.
"This allows owners of e.g. large public car parks, airports, shopping centers and private companies to monitor their car park and to offer better services to their customers. They´ll be able to automatically lead their clients to the nearest free parking space," says Jonas Hammershøj.
Although technically mature and ready for use, certain legal hurdles remain, before the concept can fully enter the Scandinavian market. A key issue is how the footage should be stored and whether certain parts of the car parks need to be blurred out or not.