FlexPay: Towards Cloud Based Automatic Number Plate Recognition System

Automatic number plate recognition (ANPR) can be utilized in many fields, including multi-lane free flow (MLFF), parking systems, and other vehicle-related systems, to keep track of vehicles. The need for a centralized ANPR mechanism emerges for a more sustainable and cheaper solution. We propose an IoT system that takes the image of vehicles and processes it in the cloud. Solutions involve IoT sensors and a camera that takes the image, a backend service that acts as the middleware, an ML service to gather the vehicle number, and a frontend as a dashboard for users. The system is built on top of the Google Cloud Platform (GCP), mainly Cloud Run as the servers, Pub/Sub communication, and MongoDB as the database. YOLOv8 is used for detecting vehicles in the image, and PaddleOCR is used to extract the vehicle number from the detected number plate. The capture of the image is signaled by the IR sensor and captured by the ESP32-CAM, which is then transferred to the backend. The research results demonstrate that this solution is not only cost-effective but also adaptable to a wide range of systems, with some adjustments and improvements. The YOLOv8 has a really high accuracy of $99 \%$, while PaddleOCR did not do that well at only $73 \%$ accuracy.