Accept Cookies

Hello, it’s me: computer vision in parkings

Thomas Smolders
Resident Writer
Sebastiaan Van den Branden
Data Scientist

One of the themes that has been the subject of almost all our conversations in recent months is definitely "hygiene". From social distancing to the continuous washing of our hands: everything revolved around how we try to avoid touching each other. Not only do we want to touch fewer people: in the near future we will most likely also be less inclined to touch devices.

A perfect example of this are the so-called smart venues, buildings wrapped with information technology, processes and services. From elevators that are operated via voice to smart locks and smart lighting. These devices at their turn generate data, which is managed on platforms that allow users to heat, cool or illuminate the buildings even better.

In our search for a very concrete example of machines that people touch when it is maybe not necessary, we decided to focus on parking meters. For example, could parking operators recognize your license plate with computer vision, so that it can be linked to a database with your payment details? If so: how would something like that work technically?

Edge computing

Along the rise of edge computing over the last few years that enabled running machine learning models on the edge (or on local devices) came a number of machine learning applications that could be deployed where privacy is the main take-away. In the case of computer vision, for instance, we saw the possibility to serve camera feed to a machine learning model on the edge, extract the necessary information without streaming or storing that camera feed ensuring privacy compliance.

Google’s Edge TPU is the perfect example of this edge computing power. It packs enough power to serve demanding computer vision algorithms with a camera feed and make predictions in real time. In the case of the parking access management system, we could hook up a camera to the Edge TPU that films the license plates, extract the license plate number and feed it to the database.

This extraction is done with optical character recognition or OCR. This database then has your payment details linked to your license plate and the transaction can be executed automatically when entering and exiting the parking lot. Besides license plates we can see access management applications that are powered with face recognition as well.


This innovative way of working is not often used in practice. Skidata, the world market leader in large-scale access and revenue management, has it integrated in its PlateTech.Logic solution, but only a few Belgian parkings use their solution. Some parkings have contactless payment terminals, but don’t combine it with ANPR yet. Maybe the current crisis could change this!

If you’d like to discover how you can automate some steps in your customer journey so your customers have to touch less devices, we can highly recommend this webinar by our Product Manager Stephanie. See you there!

Join us for our next webinar!

In the next weeks, our experts will explore how and if we can get out of the current crisis by rebooting the system and which role technology can play.

Discover more
Join us for our next webinar!

Related stories

Get more insights

Sign up for our newsletter and stay updated on trends, events and inspiring cases.