Scholarlink

<p>Icing on transmission lines is an inevitable natural phenomenon, and different types of icing bring different degrees of harm to transmission lines, especially soft rime, hard rime, and glaze, the three types of icing need more concerning. In order to know the ice types well and further to evaluate the icing state, a classification algorithm of icing types on conductors and insulators based on histogram feature and LIBSVM is proposed in this paper. Firstly, the foreground conductor or insulator region is extracted from the SLIC super-pixel images of transmission line icing images by GrabCut algorithm. Secondly, utilizing the pixel coordinate of foreground edge to calculate the center axis of foreground region, so that the foreground region can be subdivide to several regular blocks, and then the local difference processing is performed to strength the surface texture feature on each block, from the histogram distribution characteristics of texture enhancement blocks of different ice types, an automatic classification model of icing types based on LIBSVM with histogram multi-features is constructed. The recognition accuracies of the algorithm in this paper are 96.67% and 85.3%, respectively. Soft rime, hard rime, and glaze can be effectively distinguished, providing reliable status data for monitoring the icing conditions of transmission lines.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A classification method of ice types on conductors and insulators by image recognition

  • 期刊论文

  • 发表日期：

    2025年4月22日

查看全文
  • Ye Zhang,
  • Yufeng Zhan,
  • Jinghao Shang,
  • Pengchao Zhai,
  • Xinbo Huang

摘要

Icing on transmission lines is an inevitable natural phenomenon, and different types of icing bring different degrees of harm to transmission lines, especially soft rime, hard rime, and glaze, the three types of icing need more concerning. In order to know the ice types well and further to evaluate the icing state, a classification algorithm of icing types on conductors and insulators based on histogram feature and LIBSVM is proposed in this paper. Firstly, the foreground conductor or insulator region is extracted from the SLIC super-pixel images of transmission line icing images by GrabCut algorithm. Secondly, utilizing the pixel coordinate of foreground edge to calculate the center axis of foreground region, so that the foreground region can be subdivide to several regular blocks, and then the local difference processing is performed to strength the surface texture feature on each block, from the histogram distribution characteristics of texture enhancement blocks of different ice types, an automatic classification model of icing types based on LIBSVM with histogram multi-features is constructed. The recognition accuracies of the algorithm in this paper are 96.67% and 85.3%, respectively. Soft rime, hard rime, and glaze can be effectively distinguished, providing reliable status data for monitoring the icing conditions of transmission lines.