基于改进YOLOv11n的轻量级电力设备过热故障红外图像检测算法  

The Lightweight Power Equipment Overheating Fault Infrared Images Detection Algorithm Based on the Improved YOLOv11n

作　　者：周运磊 董效杰 刘三军 刘承毅 ZHOU Yunlei;DONG Xiaojie;LIU Sanjun;LIU Chengyi(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi 445000,China)

机构地区：[1]湖北民族大学智能科学与工程学院,湖北恩施445000

出　　处：《湖北民族大学学报(自然科学版)》2025年第1期114-118,140,共6页Journal of Hubei Minzu University:Natural Science Edition

基　　金：国家自然科学基金项目(61961016)。

摘　　要：针对背景复杂的变电站电力设备过热故障红外图像难以检测的问题,提出了改进你只看一次11纳米版(you only look once version 11 nano, YOLOv11n)算法。首先,采用轻量级跨尺度特征融合模块(cross-scale feature fusion module, CCFM)改进原有颈部网络,以实现对特征通道信息的高效整合并降低参数量;其次,引入具有可切换空洞卷积的2次跨阶段3卷积可变核(cross stage partial with three-convolution blocks of variable kernel size two-switchable atrous convolution, C3k2-SAConv)模块替换整个网络的C3k2模块,提升了算法的特征提取能力;最后,使用具有双层路由注意力视觉转换器的跨阶段双卷积(cross stage partial with two convolutions and vision transformer of bi-level routing attention, C2BF)模块替换跨阶段双卷积逐点空间注意力(cross stage partial with two convolutions and pointwise spatial attention, C2PSA)模块,提升了算法在复杂环境下对红外图像的检测准确度。结果表明,相较于原始YOLOv11n算法,改进YOLOv11n算法的参数量减少了22.1%;精确率、召回率、平均精确率均值分别达到91.1%、85.5%、90.9%,各自提升了3.0、2.6、2.8个百分点;检测速度达到128.2帧/s。改进YOLOv11n算法能实现对电力设备过热故障红外图像的有效检测,可满足算法轻量化与实时性检测的要求。To address the issue of detecting overheating faults in power equipment due to the complex background in infrared images of substations,an improved you only look once version 11 nano(YOLOv11n)algorithm was proposed.Firstly,the original neck network was improved by adopting a lightweight cross-scale feature fusion module(CCFM)to achieve efficient integration of feature channel information and reduce the model′s parameter quantity.Secondly,a cross stage partial with three-convolution blocks of variable kernel size two-switchable atrous convolution(C3k2-SAConv)module was introduced to replace the C3k2 module across the entire network,which enhanced the model′s feature extraction capability.Finally,a cross stage partial with two convolutions and vision transformer of bi-level routing attention(C2BF)module was used to replace the cross stage partial with two convolutions and pointwise spatial attention(C2PSA)module,which improved the model′s accuracy in detecting targets in infrared images under complex environments.The results showed that compared to the original YOLOv11n algorithm,the improved YOLOv11n algorithm parameter quantity was reduced by 22.1%;precision,recall,and mean average precision reached 91.1%,85.5%,and 90.9%,respectively,with improvements of 3.0,2.6 and 2.8 percentage point;the detection speed reached 128.2 frames/s.The improved YOLOv11n quantity was able to effectively detect overheating faults of power equipment in infrared images,meeting the requirements for lightweight and real-time detection.

关 键 词：红外图像 YOLOv11n 过热故障检测 CCFM C3k2-SAConv C2BF 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]