<![CDATA[The Ministry of Transportation reported a 4.30% increase in the number of motorized vehicles in Indonesia in 2021, making the Automatic Number Plate Recognition (ANPR) system increasingly important for traffic management. However, implementing ANPR in different weather conditions is challenging. To address this issue, a study used two deep learning modules, YOLOv5n for license plate detection and the TPS-ResNet-BiLSTM-Attn framework for character recognition. Each module was trained with two types of datasets, Dataset 1, which included images with variations in sunny and cloudy weather conditions, and Dataset 2, which included images with variations in sunny, cloudy, and moderate rainy weather conditions. The best-performing training method for the YOLOv5n model was using Dataset 2 and evolution hyperparameters, with a testing result of mAP 0.893 and f1-score 0.887. The best-performing training method for the TRBA framework was using Dataset 2 (3200 data), with a testing result of 83.08% accuracy. The ANPR system has various applications in sectors such as command forces, parking management, and road safety. The combination of object detection and character recognition allows for the development of an end-to-end AI solution for automatic license plate recognition]]>
Copyrights © 2024
