DETEKSI PELANGGARAN LALU LINTAS BERBASIS SEATBELT PADA PENGENDARA MOBIL MELALUI CCTV MENGGUNAKAN METODE YOU ONLY LOOK ONCE (YOLOv7)

Rahman Nuruddin, Abdul (2025) DETEKSI PELANGGARAN LALU LINTAS BERBASIS SEATBELT PADA PENGENDARA MOBIL MELALUI CCTV MENGGUNAKAN METODE YOU ONLY LOOK ONCE (YOLOv7). Undergraduate thesis, UPN Veteran Jawa Timur.

	Text Cover.pdf Download (1MB)
	Text Bab 1.pdf Download (98kB)
	Text Bab 2.pdf Restricted to Repository staff only until 21 May 2027. Download (820kB) \| Request a copy
	Text Bab 3.pdf Restricted to Repository staff only until 21 May 2027. Download (1MB) \| Request a copy
	Text Bab 4.pdf Restricted to Repository staff only until 21 May 2027. Download (2MB) \| Request a copy
	Text Bab 5.pdf Download (13kB)
	Text Daftar Pustaka.pdf Download (97kB)

Abstract

Driving safety is one of the main focuses in enforcing traffic regulations, particularly regarding the use of seatbelts by four-wheeled vehicle drivers. This study aims to develop an automatic seatbelt detection system using the You Only Look Once version 7 (YOLOv7) method, utilizing video recordings from a personal camera positioned on top of a flyover bridge to simulate the top-down perspective of traffic CCTV. The dataset used in this study was obtained from video recordings captured by a personal camera installed above flyovers at several roads in Surabaya. These videos were then converted into images and underwent preprocessing, manual annotation, augmentation, and model training using YOLOv7. The system was evaluated by comparing the performance of YOLOv7 under two dataset split scenarios: 80% training and 20% validation, and 75% training and 25% validation. The results showed that the 80:20 ratio yielded the best detection performance, with a mean Average Precision (mAP) of 96.4%, precision of 94.9%, and recall of 94.9%, slightly outperforming the 75:25 ratio which achieved an mAP of 96.3%, precision of 95.9%, and recall of 94.3%. These findings indicate that the 80:20 dataset ratio is more effective in improving seatbelt violation detection accuracy, especially under various lighting conditions and camera angles. The developed system is expected to serve as a solution to support automatic traffic law enforcement and contribute to increasing driver awareness and road safety.

Item Type:

Thesis (Undergraduate)

Contributors: