Perbandingan Kinerja Algoritma YOLO Dan RCNN Pada Deteksi Plat Nomor Kendaraan

Abstract

Vehicle license plate detection is a system used to automatically recognize and identify license plates on passing vehicles. This system is useful to assist in managing traffic, reducing crime, and increasing the efficiency of the transportation system. However, in its implementation, license plate detection often encounters difficulties due to variations in shape, color, and lighting conditions. In addition, vehicle license plates are also cut off or distorted in digital images, which can complicate the detection process. To solve these problems, an effective method is needed to identify objects from digital images. This study aims to evaluate and compare the performance of two popular algorithms used for vehicle license plate detection, namely YOLO (You only look once) and RCNN (Region Convolutional Neural network). The performance evaluation of the two algorithms uses the same dataset so that their accuracy can be compared. The results showed that the YOLOv4 algorithm has a higher level of detection accuracy than RCNN with an accuracy of 96% and 87,8%, respectively. Based on these results it can be concluded that YOLOv4 is more suitable for use in vehicle license plate detection applications with prioritized speeds.