面向轻量化网络的安全帽佩戴检测方法研究

Research on Safety Helmet Wearing Detection Method Oriented to Lightweight Networks

针对当前安全帽检测算法存在的普遍适用性和效率不高的问题,提出了一种基于深度学习的视觉检测算法,旨在提高建筑工地人员的安全防护水平。该算法可以自动识别并检测工地上工人是否佩戴安全帽,以实现更高效的安全管理。首先,基于当前流行的YOLO网络构建了高效提取不同环境下安全帽的特征和位置信息的安全帽佩戴检测网络。其次,为了增强实用性,本文基于PyQt5设计了工地安全帽检测的可视化操作界面。实验结果表明提出的安全帽检测模型在各种工地环境下均展现出卓越性能,能够达到92.5%的平均检测精度。这不仅实现了高精准度的安全帽自动检测,还通过可视化操作界面,极大地便利了安全帽的检测工作,充分满足了检测任务的准确性与实时性要求。

In response to the prevailing issues of limited universal applicability and suboptimal efficiency in current helmet detection algorithms,this paper introduces a vision-based detection method leveraging deep learning techniques,aiming to enhance the safety protection of construction site personnel.This algorithm is capable of automatically identifying and detecting whether workers on-site are wearing helmets,thereby promoting more efficient safety management.Initially,a helmet-wearing detection network is constructed,grounded on the prevalent YOLO network,which efficiently extracts the features and positional information of helmets in varying environments.Furthermore,to enhance usability,a visual operation interface for helmet detection on construction sites is designed utilizing PyQt5.Experimental results demonstrate that the proposed helmet detection model exhibits remarkable performance in diverse construction site environments,achieving an average detection accuracy of 92.5%.This not only realizes highly accurate automated helmet detection but also significantly facilitates helmet detection efforts through the visual operation interface,fully satisfying the accuracy and real-time requirements of the detection task.