Back to all algorithms

Please note: The algorithm descriptions in English have been automatically translated. Errors may have been introduced in this process. For the original descriptions, go to the Dutch version of the Algorithm Register.

Parking control

Municipality of Amsterdam

To keep Amsterdam liveable and accessible, only a limited number of cars are allowed to park in the city.

Last change on 25th of April 2024, at 12:25 (CET) | Publication Standard 1.0
Publication category
Impactful algorithms
Impact assessment
Field not filled in.
Status
In use

General information

Theme

  • Economy
  • Space and Infrastructure

Begin date

Field not filled in.

Contact information

Algoritmen@amsterdam.nl

Link to publication website

https://algoritmeregister.amsterdam.nl/parkeercontrole/

Responsible use

Goal and impact

To keep Amsterdam liveable and accessible, only a limited number of cars are allowed to park in the city. The municipality checks whether a parked car has the right to be parked, i.e. whether someone has paid a parking fee or has a parking permit. To work more efficiently, we do this check with scanning cars. With these, we currently check more than 150,000 official parking spaces in Amsterdam. The scan cars are equipped with cameras and drive past parking spaces in the city. The cameras capture car license plates and use an image recognition algorithm to identify license plates. The license plate number is sent to the National Parking Register where it checks whether the car has the right to be parked. If a person is found not to have paid a parking fee, a controller assesses whether there is a special situation (loading or unloading, stationary cars in front of a traffic light) based on ambient photos. If in doubt, a parking inspector reviews the situation on the spot. If there is no special situation, a parking ticket ('additional parking tax assessment') follows. Increasing the chance of being caught increases fairness towards those who do pay or have a valid parking permit. Enforcing the offence of 'standing still on the pavement' reduces nuisance and increases liveability for those who do pay or have a valid parking permit.

Considerations

Field not filled in.

Human intervention

Human controllers review the images to see if there is a special situation, for example loading or unloading, or stationary cars at a traffic light. They also check whether licence plate number is read correctly. If in doubt, a parking controller reviews the situation on the spot.

Risk management

The system has a low risk. The main risk is that the system mis-recognises a license plate and someone is fined that does not deserve it. This can happen if a letter or number on the license plate is misrecognised by both the algorithm and the controller. To manage this risk, people are given the opportunity to object in writing within six weeks via a website (naheffingsaanslag.amsterdam.nl). Those who object will be given the opportunity to see the licence plate photo and a situational photograph, if available. Any bystanders, unrelated license plates and other privacy-sensitive information will be made unrecognisable in those photos.

Operations

Data

Scanned license plates Scan cars with automated license plate recognition drive around the city to recognise license plates of parked cars. The data processed by the scan cars are the scanned images of number plates, along with the location of the car and time stamp data. If parking fees have been paid, the data is anonymised after 48 hours (license plate images and surrounding photos are deleted; the license plate is removed from the metadata). If no parking fee has been paid, data are kept for as long as necessary for objection and appeal purposes. National Parking Register (NPR) The National Parking Register is a national database in which all current parking rights of license plates are registered. The register is managed by the RDW. The data is collected from parking meters and parking apps of municipal and commercial parking companies, parking providers and enforcers. Data on parking rights are kept in the register for 13 weeks after the parking right ends (accessible only to authorised personnel). Registry data is used to check whether someone is allowed to park somewhere.

Technical design

Architecture of the model Algorithms are used: to locate the number plate in the camera images; to adjust the images for identification and; to identify the individual characters of the number plate. Once a number plate is identified and processed, the license plate data is sent to the National Parking Register for further processing. An NPR algorithm then checks the validity of parking rights for the license plate at a given time and location (for information on that algorithm, see the information on NPR's website: https://nationaalparkeerregister.nl/downloads/downloads-mobiel-parkeren.html - in particular the technical documentation). A positive result means that the car has valid parking rights. In that case, the scan data of the number plate is deleted within 48 hours. In case of a negative result, the images are checked by an employee of EGIS Parking Services B.V., which carries out parking checks on behalf of the municipality of Amsterdam. EGIS assesses whether there is a special situation, such as loading and unloading or a traffic light. If in doubt, a parking inspector looks at the situation on the spot. If a license plate has no valid parking rights, the data are transferred to the municipal tax authority. They use the RDW's database to link the license plate to the owner's personal data and send a parking ticket ('parking tax fine'). In addition to checking for paid parking and enforcing parking on the pavement, the scan car is used to detect stolen vehicles and vehicles with a police or public prosecutor's demand based on the scanned license plates. Because the scanning car runs the scanned license plates against the public data at the RDW (vehicles registered as stolen), the National Vehicle Crime Information Centre (possibly together with police) can trace that vehicle if there is a match. Data are also collected on parking pressure and the type of permit holder.