基于改进SSD的变压器套管红外图像油位智能识别方法
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

TM854

基金项目:

江苏省自然科学基金资助项目(SBK2021020744)


Intelligent oil level recognition of transformer bushing infrared image based on improved SSD algorithm
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:

    为解决现有基于红外图像识别变压器套管油位存在的过于依赖温度信息、人工处理效率低下等问题,文中结合目标检测技术提出了一种基于改进单次检测器(SSD)的套管智能油位识别方法。通过引入SSD目标检测方法,检测红外图像中的套管区域,加入损失函数以改进SSD算法从而提高套管检测准确率,并进一步通过简单线性迭代聚类(SLIC)的应用实现了不依赖红外图像温度信息的油位检测。对比文中提出的基于红外图像的油位识别算法检测结果与人工油位检测结果,表明文中提出的算法不仅在效率上领先于传统的温度检测方式,且其误差较小,仅为0.08%。对比结果验证了所提算法在保证检测精度的情况下可大幅度提高检测效率,有效提升套管故障诊断效率和智能化水平。

    Abstract:

    Recognition of transformer bushing oil level based on infrared image suffers from over-reliance on temperature information and low efficiency on manual processing. In order to solve the existing problems,an intelligent bushing oil level recognition method based on improved single shot detection (SSD) with the combination of target detection technique is proposed in this paper. The bushing area in the infrared image is detected through introducing the SSD target algorithm. And SSD algorithm is improved by adding the center loss function. Furthermore,oil level detection of bushing is achieved through the application of simple linear iterative clustering (SLIC) without relying on the temperature information. Comparative results among the image-based oil level recognition algorithm proposed in the paper and the manual oil level detection method,show that the efficiency of the proposed one is prior to that of the traditional temperature-based method. Moreover,the relative error of the algorithm is only 0.08%. Therefore,the proposed algorithm greatly improves the detection efficiency while ensuring the detection accuracy,so as to enhance the efficiency of bushing fault diagnosis and the degree of intelligence.

    参考文献
    相似文献
    引证文献
引用本文

别一凡,李波,江军,张潮海.基于改进SSD的变压器套管红外图像油位智能识别方法[J].电力工程技术,2021,40(5):158-163

复制
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:2021-03-22
  • 最后修改日期:2021-05-28
  • 录用日期:2020-10-19
  • 在线发布日期: 2021-09-30
  • 出版日期: 2021-09-28
文章二维码