摘要
文章探讨了建筑工程施工现场中不安全行为的监测与识别技术,重点设计了一种结合摄像头优化布置与无人机图像采集的解决方案,通过CNN-LSTM和YOLOv5算法对不安全行为进行了识别与检测。通过实验验证,CNN-LSTM在不安全动作识别上达到95.8%的准确率,而YOLOv5在安全物品佩戴检测中则实现了97.2%的准确率。这些技术的集成显著提高了施工现场的安全管理效率。
This study discusses the monitoring and identification technology of unsafe behaviors in the construction site of construction engineering,focuses on designing a solution combining the optimization of camera layout and UAV image acquisition,and identifies and detects unsafe behaviors through CNN-LSTM and YOLOv5 algorithm.Through experimental verification,CNN-LSTM achieved 95.8%accuracy in the identification of unsafe actions,while YOLOv5 achieved 97.2%accuracy in the wearing detection of safe items.The integration of these technologies significantly improves the safety management efficiency of the construction site.
作者
杨佩锋
YANG Peifeng(Gansu Sixth Construction Group Co.,Ltd.,Lanzhou 730030,China)
