| Summary: | Object detection Car Plate Recognition (CPR), Automatic Number Plate Recognition (ANPR), and Automatic License Plate Recognition (ALPR) are smart technologies that can detect and recognize the vehicle number plate to create location data with the help of scanners or cameras. The rapidly increased number of vehicles in Malaysia leads to traffic congestion and higher traffic violation problems. Hence, the capabilities of the CPR system to automatically recognize vehicle plate numbers would be a game-changing situation. It would be a key step to reduce and improve the legislation on traffic and traffic congestion problems. In this study, 205 real-time images of Vietnamese number plates were used in the training process. For this research, YOLOv4 architecture was selected and designedfor the number plate detection, and a pre-trained CNN model was used for the character recognition using Python 3, Ten sorflo w, Keras library, and YOLO-Darknet. Two experiments were conducted to find the best YOLOv4 parameters' settings using different learning rate values and batch sizes. The results obtained for both experiments were satisfactory, where the model with a learning rate of 0.001, batch size of 64, and 90:10 splitting ratio had the best performance, with a value of mAP of 1 and an average IoU of 91.32%. The study could be further enhanced with different datasets and algorithms. © 2024 IEEE.
|
|---|