Penerapan Algoritma YOLOv8 dalam Deteksi Penggunaan Helm dan Plat Nomor Polisi Kendaraan pada Sepeda Motor Berbasis Android

Abstract

The number of motorcycle accidents and injuries has increased by 6.8% compared to 2022. The Central Bureau of Statistics (BPS) Indonesia has recorded that most fatal accidents involving motorcycle riders are due to inadequate helmet use or no helmet use at all. Therefore, a system has been developed to detect helmet violations in real-time and automatically record vehicle license plate numbers. This research employs the You Only Look Once version 8 (YOLOv8) method and utilizes EasyOCR. YOLOv8 is capable of detecting helmet objects and vehicle license plates on motorcycles. EasyOCR is used for extracting vehicle license plate numbers. The testing results of the You Only Look Once version 8 algorithm model achieved an accuracy rate of 96% with an average precision of 96.5%, recall of 100%, and an F1-score of 98% in detecting helmet use among riders. In detecting vehicle license plate numbers, it achieved an accuracy rate of 89.5% with an average precision of 98%, recall of 91%, and an F1-score of 94%. With the obtained evaluation metrics, the detection system for identifying helmet users and vehicle license plates on motorcycles performs exceptionally well.