摘要
针对电网施工环境多样、传统方法存在监控盲区导致监控精度和覆盖面低的问题,提出了一种面向电网工作人员识别的轻量化检测网络,以高空及地面立体化的巡查模式来监控电网工作人员的作业行为。该方法采用一个轻量化的目标检测网络检测出监控视频中的电网工作人员,并判断其是否佩戴安全帽,使用人员识别网络来辨别未佩戴安全帽人员的身份。仿真实验结果表明,所提方法可以实现立体化电网工作人员作业行为巡查,相比于传统方法,所提出的轻量化网络具有更小的计算量,可达到63.4%的识别精度。
Aiming at the problems of low monitoring accuracy and coverage caused by the diversity of power grid construction environment and the existence of monitoring blind areas in traditional methods,a lightweight detection network dedicated to the identification of power grid workers was proposed to monitor the work behavior of power grid workers through the high-altitude and ground three-dimensional patrol mode.A lightweight target detection network was used to detect the power grid workers in the surveillance video,and judge whether they wore helmets or not,and then the personal recognition network was used to identify the worker not wearing helmets.The simulation results show that the as-proposed method can realize the three-dimensional inspection of the work behavior of power grid workers.Compared with the traditional method,the as-proposed method has lesser computation and can achieve a recognition accuracy of 63.4%.
作者
胡戈飚
林志驰
郭政
赵文硕
HU Gebiao;LIN Zhichi;GUO Zheng;ZHAO Wenshuo(School of Electrical Engineering,Nanchang University,Nanchang 330036,Jiangxi,China;Branch of State Grid Jiangxi Electric Power Co.,Ltd.,Nanchang 330012,Jiangxi,China;School of Electrical Engineering and Electronic Information,Xihua University,Chengdu 610000,China)
出处
《沈阳工业大学学报》
CAS
北大核心
2024年第3期248-254,共7页
Journal of Shenyang University of Technology
基金
国家自然科学基金项目(51367014)
国家电网公司科技项目(52182420001B)。
关键词
监控视频
金字塔池化网络
安全帽佩戴
深度学习
轻量化
身份识别
surveillance video
pyramid pooling network
safety helmet wearing
deep learning
lightweight
personal recognition
