Safety Helmet Detection System of Smart Construction Site Based on YOLOv5S
Author/Presenter: Zhou, Minghang; Zhu, Jiuyu; Li, XinyuAbstract:
Because of the high risk of construction work, the safety of construction personnel have been widely paid attention from all walks of life. Some construction personnel does not obey the rules and regulations of the construction site, such as often do not wear safety helmets, which greatly increases the potential safety hazard (on the construction site). However, since the construction site environment is relatively complex and usually of the small target, with relatively backward hardware facilities and low economic investment, the model real-time detection performance is not ideal and has low robustness. To settle these problems, this paper proposes a YOLOv5 model to judge the real-time proportion of work-site construction workers wearing safety helmets and build the helmet detection for the complex environment based on YOLOv5S. Experimental evaluation shows that the verification of the given safety helmet wear test data set reached the average accuracy of 93.8%. Compared with the previous YOLOv5 algorithm, the accuracy of helmet-wearing detection is improved by 1.1%, which meets the helmet detection requirements in complex construction scenes.
Publication Date: 2022
Source URL: Link to URL
Publication Types: Books, Reports, Papers, and Research Articles
Topics: Detection and Identification; Hard Hats; Machine Learning; Worker Safety