可提高检测精度与速度的电力安全带智能检测方法  

Intelligent Detection Method for Power Safety Belts Improving Detection Accuracy and Speed

作　　者：朱建宝 桑顺[2] 邓伟超 马青山 张斌 圣勇 Zhu Jianbao;Sang Shun;Deng Weichao;Ma Qingshan;Zhang Bin;Sheng Yong(Nantong Power Supply Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Nantong Jiangsu 226006,China;Key Laboratory of Control of Power Transmission and Conversion,Shanghai Jiao Tong University,Shanghai 200240,China;Nantong Aowei Information System Engineering Co.,Ltd.,Nantong Jiangsu 226007,China)

机构地区：[1]国网江苏省电力有限公司南通供电分公司,江苏南通226006 [2]上海交通大学电力传输与功率变换控制教育部重点实验室,上海200240 [3]江苏奥威信息系统工程有限公司,江苏南通226007

出　　处：《电气自动化》2025年第1期20-22,共3页Electrical Automation

基　　金：电力传输功率变换控制教育部重点实验室开放课题(2021AC03);国网江苏省电力有限公司科技项目(J2020054)。

摘　　要：针对室外电力检修场景成像干扰大、图像背景复杂、电网检修作业人员未能及时正确佩戴安全带的问题,提出了一种电力安全带的智能检测方法。首先,利用均值聚类训练目标框来调整网络的锚框,使网络获得检测目标的先验知识;然后,在主干网络中提取更加丰富的特征图,使用广义交并比损失代替原始边框的回归损失;最后,通过预测层直接回归目标类别。研究结果表明,相较常规检测方法,采用所提方法平均检测精度为83.15%,且检测速度可达到每秒62帧,兼有更高的检测精度和更快的检测速度。A smart detection method for power safety belts was proposed to address the issues of high imaging interference,complex image backgrounds,and failure of power grid maintenance workers to wear safety belts in a timely manner in outdoor power maintenance scenarios.Firstly,using mean clustering to train the target box and adjust the anchor box of the network,the network obtained prior knowledge of the detected target;then,richer feature maps were extracted from the backbone network,and the generalized intersection union ratio loss was used to replace the regression loss of the original bounding box;finally,the target category was directly regressed through the prediction layer.The research results show that compared to conventional detection methods,the detection accuracy of the proposed method is 83.15%,and the detection speed can reach 62 frames per second,with higher detection accuracy and faster detection speed.

关 键 词：电力安全带 深度学习 目标检测 检测精度 检测速度 

分 类 号：TM715[电气工程—电力系统及自动化]