News
AI wants to fine-tune metropolitan ZBE licence plate reading by 2028
ATM has awarded EYSA a contract of 156,594 euros to prepare the metropolitan ZBE system. Licence plate reading must reach 97.5% accuracy and reduce errors by 60%.
Night-time and rain make automatic number plate recognition more difficult. Challenging angles also test the cameras: the new ZBE system will need to improve its accuracy before restrictions are expanded from 2028.
Fewer errors in identifying vehicles
ATM has awarded Estacionamientos y Servicios, S.A.U. (EYSA) the management and improvement of the IT system for the next two years. The contract is worth 156,594 euros excluding VAT and includes the possibility of two additional extensions.
Optical character recognition (OCR), which identifies number plates captured by the cameras, will need to achieve a minimum accuracy of 97.5%, compared to the 94.2% indicated for the current system. There is also a requirement to reduce false positives and negatives by 60% through machine learning.
Available figures show that the system fails to detect around 3.3% of vehicles and records incorrect readings in another 2.5% of cases. The aim is to reduce these errors when images are captured in low light or rain, among other challenging conditions.
Up to 20 municipalities and restrictions planned for 2028
From 2028, EYSA will process and analyse images sent by cameras located outside the metropolitan area. The infrastructure is being prepared for the planned extension of restrictions to vehicles with environmental label B in municipalities with more than 50,000 inhabitants.
The service may also cover towns with more than 20,000 inhabitants that exceed legal pollution limits. In total, the planned scope covers up to 20 municipalities, as previously explained in information about EYSA.