<![CDATA[Traffic violations committed by motorcycle riders remain a significant challenge in Indonesia’s transportation sector and contribute to an increased risk of traffic accidents. Common violations include riding without helmets and carrying more than one passenger. This study proposes a motorcycle traffic violation detection approach using the YOLOv7 object detection model applied to CCTV monitoring data. The dataset used in this study was obtained from CCTV recordings at Jalan Gapura DPRD, Bandar Lampung, and consists of images representing three traffic density scenarios: low, medium, and high. The research methodology includes frame extraction, image preprocessing, object labeling, model training, testing, and performance evaluation using standard object detection metrics, namely Precision, Recall, and mean Average Precision (mAP). Experimental results show that the YOLOv7 model achieves consistently high detection performance across different traffic density conditions. The best performance was obtained using a learning rate of 0.01 with 100 training epochs, achieving 99.66% mAP in low traffic conditions, 98.75% in medium traffic conditions, and 97.08% in high traffic conditions. These results indicate that the proposed approach has strong potential for implementation in CCTV-based traffic monitoring systems and may support automated traffic violation detection as part of Electronic Traffic Law Enforcement (ETLE) initiatives.]]>
Copyrights © 2026
