基于改进YOLOv7的电力设备红外过热缺陷检测方法  

Infrared overheating defect detection method for power equipment based on improved YOLOv7

作　　者：林丽霞 吴悦园 LIN Lixia;WU Yueyuan(Zhanjiang Power Supply Bureau of Guangdong Power Grid Co.,Ltd,Zhanjiang,Guangdong 524000)

机构地区：[1]广东电网有限责任公司湛江供电局,广东湛江524000

出　　处：《电气技术》2024年第1期42-47,共6页Electrical Engineering

摘　　要：电力设备运行时发生过热缺陷容易引起电气故障,严重威胁电力设备安全运行。为了有效监测电力设备运行状态,提出一种基于改进YOLOv7的电力设备红外过热缺陷检测方法。采用YOLOv7目标检测网络作为基础检测网络,使用CIoU衡量矩形框的损失,同时将原网络的空间金字塔池化-跨阶段局部网络(SPPCSPC)结构替换为快速空间金字塔池化-跨阶段局部网络(SPPFCSPC)结构对模型进行改进,以增大模型感受野并提升对过热缺陷的检测性能。实验结果表明,基于改进YOLOv7的检测方法的准确率达到90.39%、召回率达到78.95%、平均正确率达到89.64%,可为电力设备过热缺陷红外检测提供参考。Overheating defects in power equipment during operation can easily cause electrical faults,posing a serious threat to the safe operation of power equipment.In order to effectively monitor the operation status of power equipment,a method for detecting infrared overheating defects in power equipment based on improved you only look once v7(YOLOv7)is proposed.YOLOv7 object detection network is used as the basic detection network,and the loss of rectangular boxes is measured by using complete intersection over union(CIoU).At the same time,the spatial pyramid pooling-cross-stage partial channel(SPPCSPC)structure is replaced by the spatial pyramid pooling-fast-cross-stage partial channel(SPPFCSPC)structure of the original network to improve the model,while increasing the receptive field of the model and improving the overheating defects detection performance.The experimental results show that the precision rate of this method based on improved YOLOv7 reaches 90.39%,the recall rate reaches 78.95%,and the average precision value reaches 89.64%,which can provide technical reference for infrared detection of overheating defects in power equipment.

关 键 词：电力设备 红外图像 缺陷检测 改进YOLOv7 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术] TM50[电气工程—电器]