YOLOv8-Based Real-Time Pedestrian Detection Model for Urban Traffic Surveillance Systems

Le résumé fourni par la source

Pedestrian detection plays an important role in improving road safety and enabling intelligent traffic monitoring in modern urban environments. This study investigates the use of deep learning techniques for real-time pedestrian detection from urban traffic video streams. In particular, a Pedestrian Detection YOLOv8 (PD-YOLOv8) model is developed to automatically identify pedestrians in captured surveillance footage. The model is trained using the City Street View training dataset, which comprises 2,528 urban street images extracted from traffic video recordings. In addition, a web-based urban traffic surveillance system was developed to demonstrate real-time pedestrian detection from video streams at the Bukit Bintang MRT junction. Experimental results demonstrate that the PD-YOLOv8 model achieves an improved detection performance of 79.65% precision, 76.07% recall, and an F1-score of 77.82%, with mAP@0.5 of 82.40% and mAP@0.5:0.95 of 64.80% as compared with baseline YOLOv8 and YOLOv5 models. These results can support pedestrian monitoring within urban traffic surveillance systems, contributing to safer road conditions and providing new features for more advanced intelligent transportation systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.