摘要
变电站现场作业中,确保电力工作人员规范佩戴安全帽的研究具有非常重要的意义。现有变电站监视安全帽佩戴情况的方式主要依靠人工看视频进行判断,效率及准确率低。本文提出基于Pytorch框架对深度神经网络架构进行迁移学习,通过算法实现自动识别安全帽佩戴状态的功能,并创新采用Focal损失函数对分类层进行安全帽佩戴识别以增强最终分类预测的准确性。实验结果表明,该改进网络具有良好性能,并在变电站现场应用中具有有效性。
In the field operation of substation,it is of great significance to ensure the power workers.The existing substation to monitor the wearing of safety helmet mainly depends on manual watching video to judge,with low efficiency and accuracy.This paper proposes the transfer learning of deep neural network architecture based on Pytorch framework,realizes the function of automatic recognition of helmet wearing status through algorithm,and innovatively uses the Focal loss function to identify the classification layer to enhance the accuracy of the final classification prediction.The experimental results show that the improved network has good performance and is effective in substation field application.
作者
张书航
张桥良
邬鲁明
张书婷
邹攀峰
周雪雪
ZHANG Shuhang;ZHANG Qiaoliang;WU Luming;ZHANG Shuting;ZOU Panfeng;ZHOU Xuexue(School of Electric Power Engineering,Nanjing University of Technology,Nanjing 211100,Jiangsu,China;Zhoushan Qiming Power Supply Service Co.,Ltd.,Zhoushan 316000,Zhejiang,China;Ningbo Vocational and Technical College,Ningbo 315800,Zhejiang,China;Putuo Mountain Guanyin Law Management Center,Zhoushan 316000,Zhejiang,China;State Grid Zhoushan Power Supply Company,Zhoushan 316000,Zhejiang,China)
出处
《流体测量与控制》
2025年第6期18-22,共5页
Fluid Measurement & Control
关键词
安全帽识别
变电站
神经网络
图像识别
safety helmet recognition
substation
neural network
image recognition
