Scholarlink

Power-line insulators play a vital role in electrical transmission systems, and precise defect detection is crucial for guaranteeing their optimal performance, maintaining grid reliability, and ensuring public safety. This study investigates and compares the effectiveness of three feature extraction methods—Local Binary Pattern (LBP), Grey Level Co-occurrence Matrix (GLCM), and Local Directional Pattern (LDP) for detecting defects in power-line insulators, utilizing the K-Nearest Neighbors (KNN) classifier. This investigation aims to assess the efficiency of each method, focusing on accuracy, precision, and overall classification efficacy, to determine their relative strengths and limitations. Experimental results show an accuracy of 96.5%, compared to 80.4% and 70.6%, respectively. LBP's robustness to illumination variations and computational efficiency render it ideal for real-time inspection. The study presents key findings that inform the selection of feature extraction methods for defect detection, contributing to the optimization of defect detection and thereby enhancing grid reliability and safety.

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

Comparison of Feature Extraction Methods for Power-Line Insulator Defect Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Osatohanmwen Noghayin Enehizena,
  • Osemwegie S. Omorogiuwa,
  • Solomon Ogiobadan,
  • Agbotiname Lucky Imoize,
  • Youssef Mejdoub

摘要

Power-line insulators play a vital role in electrical transmission systems, and precise defect detection is crucial for guaranteeing their optimal performance, maintaining grid reliability, and ensuring public safety. This study investigates and compares the effectiveness of three feature extraction methods—Local Binary Pattern (LBP), Grey Level Co-occurrence Matrix (GLCM), and Local Directional Pattern (LDP) for detecting defects in power-line insulators, utilizing the K-Nearest Neighbors (KNN) classifier. This investigation aims to assess the efficiency of each method, focusing on accuracy, precision, and overall classification efficacy, to determine their relative strengths and limitations. Experimental results show an accuracy of 96.5%, compared to 80.4% and 70.6%, respectively. LBP's robustness to illumination variations and computational efficiency render it ideal for real-time inspection. The study presents key findings that inform the selection of feature extraction methods for defect detection, contributing to the optimization of defect detection and thereby enhancing grid reliability and safety.