【摘要】 随着国内基础建设的大力发展，保证施工现场的安全成为了全社会关注的热点。为了提高施工现场监管智能化程度，建立了基于边云协同架构的施工现场安全帽检测系统，基于目标检测技术与深度学习技术，提出了一种改进Yolo V3的安全帽检测方案。针对Yolo V3算法对特定目标识别程度低，使用K-means算法得到了符合安全帽检测的先检框尺寸，并对Yolo V3的损失函数进行了改进。将改进好的模型在云端高性能GPU服务器进行训练，训练好的模型部署在边缘设备中，并配合云端ESC设备实现远程施工现场安全监控。更多还原

【Abstract】 With the vigorous development of domestic infrastructure, ensuring the safety of construction site has become the focus of the whole society. In order to improve the intelligence of construction site supervision, a construction site helmet detection system based on edge-cloud collaborative architecture is established. Based on target detection technology and deep learning technology, an improved helmet detection method for Yolo V3 is proposed. In view of the low recognition degree of Yolo V3 algorithm for specific targets, the first inspection frame size in line with helmet detection is obtained by using k-means algorithm, and the loss function of Yolo V3 is improved. The improved model is trained on the cloud high-performance GPU server, the trained model is deployed in the edge equipment, and cooperate with the cloud ESC equipment to realize remote construction site safety monitoring.更多还原