计算机科学
风险分析(工程)
组分(热力学)
可靠性工程
工程类
顺从(心理学)
系统工程
钥匙(锁)
过程(计算)
度量(数据仓库)
作者
Ashwin Sathappan Vellayan,Bala M,Bhuvan Kalyan G V,Giridharan P,G Janani
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-04-13
卷期号:: 1126-1136
标识
DOI:10.1109/icsedis68157.2026.11517930
摘要
Construction sites are highly hazardous environments where failure to adhere to Personal Protective Equipment (PPE) regulations significantly increases worker risk. Conventional manual surveillance is labor-intensive and impractical for large-scale operations, necessitating automated intelligent monitoring solutions. This study proposes a real-time multi-camera PPE compliance framework that integrates lightweight YOLO-based object detection models with ByteTrack, temporal persistence filtering, and zone-based violation analysis. The system processes multiple video streams concurrently to enable scalable safety monitoring. Four lightweight architectures - YOLOv8n, YOLOv8s, YOLOv8m, and YOLOv11n - were benchmarked using publicly available PPE datasets to identify the most suitable model for deployment. Experimental results demonstrate that YOLOv11n achieves the best accuracy-latency trade-off, attaining an mAP@50 of 88% and an average inference latency of 8.69 ms across four concurrent video streams. Furthermore, the integration of temporal filtering and contextual zone-based logic significantly reduces false-positive alerts, improving system reliability. These findings validate the effectiveness of lightweight YOLO models for robust, real-time safety monitoring in complex industrial environments.
科研通智能强力驱动
Strongly Powered by AbleSci AI