Scholarlink

<p>Aiming at the actual scenario of wind farm, this research releases a wind turbine blades surface defect dataset named WTBs2025. The images in this dataset are captured from the wind farm sites and then preprocessed for the following data augmentation. However, due to the small number of lightning strike defect samples, data augmentation based on GA-DCGAN is performed, after some simple augmentation strategies. WTBs2025 has a total of 7544 blade defect images covering 9 wind turbine blade defect types. WTBs2025 is characterized by its diverse defect types, extensive collection of defect images across multiple categories, and well-established features. These features include broad defect coverage, high-quality annotations, standardized data formats, and multi-scenario applicability. WTBs2025 demonstrates its unique practical value in the wind turbine blade damage object detection field under its high degree of compatibility and specificity. Experiments show that this dataset performs well on several well-known object detection models and provides valuable image data resources for the research and application of wind turbine blade damage detection and other directions.</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

Wind turbine blades surface defect high-quality image dataset construction and performance validation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Ruihua Zhang,
  • Sunxin Wang,
  • Sheng Liu,
  • Yu Liang,
  • Taoyu Wang,
  • Chao Liu,
  • Shu Chen

摘要

Aiming at the actual scenario of wind farm, this research releases a wind turbine blades surface defect dataset named WTBs2025. The images in this dataset are captured from the wind farm sites and then preprocessed for the following data augmentation. However, due to the small number of lightning strike defect samples, data augmentation based on GA-DCGAN is performed, after some simple augmentation strategies. WTBs2025 has a total of 7544 blade defect images covering 9 wind turbine blade defect types. WTBs2025 is characterized by its diverse defect types, extensive collection of defect images across multiple categories, and well-established features. These features include broad defect coverage, high-quality annotations, standardized data formats, and multi-scenario applicability. WTBs2025 demonstrates its unique practical value in the wind turbine blade damage object detection field under its high degree of compatibility and specificity. Experiments show that this dataset performs well on several well-known object detection models and provides valuable image data resources for the research and application of wind turbine blade damage detection and other directions.