基于改进YOLOv5接触网绝缘子定位方法
DOI:
作者:
作者单位:

华东交通大学

作者简介:

通讯作者:

中图分类号:

基金项目:

牵引供电系统全生命周期健康状态智能检测与诊断(编号:GJJ210652)


Based on the improved YOLOv5 catenary insulator positioning method
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    针对高速铁路接触网绝缘子在复杂背景下检测效率不高的问题,首先对样本数据集进行大规模扩充;在原有YOLOv5s算法的基础上,为有效的提升模型的表征力,增加ECA注意力机制,进行无降维的跨信道方式来聚焦绝缘子位置信息;使用BiFPN特征金字塔网络,进行多尺度的特征融合来丰富语义信息;选用Meta-ACON自适应控制激活函数,在函数允许的最大范围内,严格把控函数的上下限,防止模型出现失控现象;将原有GIOU损失函数更换为EIOU损失函数,从梯度的角度对锚框进行更深一步的划分,进而提升网络的收敛速度;最后,根据实验结果表明,通过对YOLOv5s改进后的检测算法,可以对绝缘子进行更精确的定位与识别,准确率达到了99.4%;所提出的检测算法为绝缘子定位检测提供了更加准确快捷的方法。

    Abstract:

    Aiming at the problem that the detection efficiency of high-speed railway catenary insulators is not high in complex background, the sample dataset is first expanded on a large scale. On the basis of the original YOLOv5s algorithm, in order to effectively improve the representation power of the model and increase the ECA attention mechanism, a cross-channel method without dimensionality reduction is carried out to focus on the position information of insulators. The BiFPN feature pyramid network is used to enrich the semantic information by multi-scale feature fusion. Select the Meta-ACON adaptive control activation function, and strictly control the upper and lower limits of the function within the maximum range allowed by the function to prevent the model from running out of control. The original GIOU loss function is replaced with the EIOU loss function, and the anchor box is further divided from the perspective of gradient, so as to improve the convergence speed of the network. Finally, according to the experimental results, the improved detection algorithm of YOLOv5s can be used to locate and identify the insulator more accurately, and the accuracy rate reaches 99.4%. The proposed detection algorithm provides a more accurate and faster method for insulator positioning detection.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2023-02-28
  • 最后修改日期:2023-03-20
  • 录用日期:2023-03-20
  • 在线发布日期: 2023-06-21
  • 出版日期: