Toll gate systems with YOLO-CNN based using Raspberry Pi for vehicle class detection

Traffic congestion at toll gates remains a significant issue, primarily due to manual vehicle identification and payment processes. This paper aims to design a prototype of an automated toll gate system integrated with IoT that can automatically detect vehicle classes with the You Only Look Once (YOLO)- Convolutional Neural Networks (CNN) based, and supports digital payment methods via RFID and QRIS. The system is built using a Raspberry Pi 3B+ as the controller, equipped with an RFID-RC522 module and an MG90S servo motor as the gate actuator. A 604 dataset of vehicle images was used to train the YOLOv3 model, achieving a mean Average Precision (mAP) of 75.41%. Experimental results show that the system can read RFID data reliably and receive QRIS payment callbacks in real time through integration with Flask and Ngrok web applications. The implementation demonstrates that all components work in an integrated and responsive performances for vehicle detection, identity verification, and gate control. This system presents strong potential as an efficient solution to reduce queues and improve traffic flow at toll gates.