电工技术2024,Issue(1) :176-179.DOI:10.19768/j.cnki.dgjs.2024.01.045

基于改进Faster R-CNN的绝缘子缺陷检测识别与定位

Insulator Defect Detection,Identification and Location Based on Modified Faster R-CNN

贺元帅 纪超 王博雅 贾星海 张凡 李小兵
电工技术2024,Issue(1) :176-179.DOI:10.19768/j.cnki.dgjs.2024.01.045

基于改进Faster R-CNN的绝缘子缺陷检测识别与定位

Insulator Defect Detection,Identification and Location Based on Modified Faster R-CNN

贺元帅 1纪超 1王博雅 1贾星海 1张凡 1李小兵1
扫码查看

作者信息

  • 1. 西安工程大学,陕西 西安 710048
  • 折叠

摘要

针对现有算法对绝缘子检测精度不高的问题,在 Faster R-CNN 算法的基础上进行改进,利用检测效果更好、性能更优的 ResNet50 代替原始 VGG 网络进行缺陷识别.实验结果表明,改进算法在数据集上的 mAP 达到77.29%,召回率达到 87.55%,与其他经典算法相比具有更好的准确性与较强的实时性.

Abstract

In view of the problem of unsatisfactory accuracy of existing algorithm in insulator detection,the present work made an improvement by using Faster R-CNN algorithm,and introduced Resnet50,which has better detection utility and excellent performance,to replace the original VGG network for defect identification.The improved algorithm,according to experimental results,can achieve a mAP reaching 77.29%on the data set in this paper,and a recall rate of 87.55%,exhibiting better accuracy and stronger real-time performance compared with other typical algorithms.

关键词

绝缘子/准确性/实时性/FasterR-CNN/Resnet网络

Key words

insulator/accuracy/real-time/Faster R-CNN/Resnet network

引用本文复制引用

出版年

2024
电工技术
重庆西南信息有限公司(原科技部西南信息中心)

电工技术

影响因子:0.177
ISSN:1002-1388
参考文献量6
段落导航相关论文