期刊文献+

基于深度学习的绝缘子自爆检测方法研究 被引量:7

Research on Insulator Self Explosion Detection Method Based on Deep Learning
下载PDF
导出
摘要 针对绝缘子自爆海量图片与人工检测效率不平衡不匹配的问题,文中提出了基于YOLOV3-tiny的绝缘子自爆自动检测方法.借鉴残差网络的跳跃连接思想,对YOLOV3-tiny模型的主干网络进行改进,提高模型特征学习能力;为了使模型更加适于绝缘子自爆检测,对Anchor box参数进行调整.实验结果表明,改进后模型的检测精确率AP达到92.57%,检测速度FPS达到38 f/s,能够快速准确地从海量航拍图像识别绝缘子自爆故障,提高了绝缘子自爆检测的自动化程度. Aiming at the problem of mismatch between the mass pictures of insulator self explosion and the efficiency of manual detection,an automatic detection method of insulator self explosion based on YOLOV3-tiny is proposed.Based on the idea of jump connection of Residual network,the backbone network of YOLOV3-tiny model is improved to improve the learning ability of model features;In order to make the model more suitable for insulator self explosion detection,the anchor box parameters are adjusted.The experimental results show that the improved model has a detection accuracy of 92.57%AP and a detection speed of 38 f/s FPS,which can quickly and accurately identify the insulator self explosion fault from a large number of aerial images,and improve the automation of insulator self explosion detection.
作者 王义军 曹培培 王雪松 闫星宇 Wang Yijun;Cao Peipei;Wang Xuesong;YanXingyu(Electrical Engineering College,Northeast Electric Power University,Jilin Jilin 132012;State Grid Tonghua Power Supply Company,Tonghua Jilin 134000;State Grid Jibei Electric Power Company.LTD.,Zhangjiakou Power Company,Zhangjiakou Hebei 075001)
出处 《东北电力大学学报》 2020年第3期33-40,共8页 Journal of Northeast Electric Power University
基金 国家自然科学基金资助项目(51877035)。
关键词 绝缘子自爆 YOLOV3-tiny 残差网络 Insulator self-explosion YOLOV3-tiny Residual Network
  • 相关文献

参考文献12

二级参考文献71

共引文献313

同被引文献69

引证文献7

二级引证文献39

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部