现代电子技术2025，Vol.48Issue(10)：76-84,9.DOI:10.16652/j.issn.1004-373x.2025.10.013
边缘计算下的绝缘子缺陷小样本检测研究
Research on insulator defect small-sample detection under edge computing
摘要
Abstract
In order to solve the problems of low precision and poor robustness of traditional target detection algorithms when detecting small-sample insulator defects on transmission lines,and to realize the efficiency of UAV patrol inspection,an edge computing based feature distance difference small-sample insulator self-explosion detection algorithm is proposed.The RT-DETR encoder is improved by means of high and low frequency information fusion(AHiLo)and(hierarchical scale-based path aggregation network,HS-PAN)to extract local high and low frequency information of insulator strings.A distance difference X-goal(DX)is introduced to obtain the optimal metric distance between the prototype agent and the query feature in the mapping feature space,so as to realize the accurate detection of small-sample insulator self-destruction.The experimental results show that the improved model can realize a detection accuracy of 86.4%using only 150 samples on the PC side,with a parameter count of 2.06×107,and a detection speed of 66.1 f/s,which meets the requirements of small-sample detection.In comparison with other mainstream algorithms,the improved algorithm can show a high level of detection accuracy and real-time performance.关键词
绝缘子缺陷/小样本检测/RT-DETR编码器/边缘计算/距离嵌入模块/路径聚合网络Key words
insulator defect/small-sample detection/RT-DETR coder/edge computing/DX module/path aggregation network分类
信息技术与安全科学引用本文复制引用
李旭涛,李宏杰,贾璐萌,邓若宇,杜剑锋,王安红..边缘计算下的绝缘子缺陷小样本检测研究[J].现代电子技术,2025,48(10):76-84,9.基金项目
国家自然科学基金项目(62072325) （62072325）
山西省研究生教育项目(2022YJJG190) （2022YJJG190）
山西省研究生实践创新项目(2024SJ319) （2024SJ319）
