Scholarlink

<p>This study proposes a two-stage machine learning framework for automated mirror detection and segmentation in concentrated solar power (CSP) systems, focusing on parabolic trough collector (PTC) and Fresnel systems. AI-based maintenance in CSP systems relies on the accurate identification and segmentation of mirrors, allowing AI models to detect and address faults. The proposed approach leverages YOLOv8 to generate accurate bounding boxes and employs the segment anything model (SAM) for detailed segmentation, thereby addressing challenges in operation and maintenance (O&amp;M). A diverse dataset of 4000 images captured at the Green Energy Park facility under various conditions, including differing sunlight, shadows, soiling, and nighttime low-light scenarios, supports the evaluation. The experimental evaluation extensively computed and analyzed results using various metrics including Precision, Recall, F1 Score, and Intersection over Union (IoU) to quantify segmentation accuracy. The integrated YOLOv8 + SAM framework reveals that approach achieves scores of up to 0.97 of IoU, representing an improvement of approximately 8–10 percentage points over YOLOv8 alone across different scenarios. Notably, a minor decrease is observed in shadowed conditions for PTC and Fresnel mirrors, primarily due to YOLO's detection challenges. The proposed framework provides a promising foundation for developing AI-driven, automated O&amp;M solutions that enhance real-time maintenance in CSP systems.</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

Advanced detection and segmentation of parabolic trough collector and Fresnel mirrors for CSP maintenance using YOLOv8 and segment anything model

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Mohamed Boujoudar,
  • Amine Moulay Taj,
  • Yasser El Jarida,
  • Ibtissam Bouarfa,
  • Ammar Mouaky,
  • Massaab El Ydrissi,
  • El Ghali Bennouna,
  • Hicham Ghennioui

摘要

This study proposes a two-stage machine learning framework for automated mirror detection and segmentation in concentrated solar power (CSP) systems, focusing on parabolic trough collector (PTC) and Fresnel systems. AI-based maintenance in CSP systems relies on the accurate identification and segmentation of mirrors, allowing AI models to detect and address faults. The proposed approach leverages YOLOv8 to generate accurate bounding boxes and employs the segment anything model (SAM) for detailed segmentation, thereby addressing challenges in operation and maintenance (O&M). A diverse dataset of 4000 images captured at the Green Energy Park facility under various conditions, including differing sunlight, shadows, soiling, and nighttime low-light scenarios, supports the evaluation. The experimental evaluation extensively computed and analyzed results using various metrics including Precision, Recall, F1 Score, and Intersection over Union (IoU) to quantify segmentation accuracy. The integrated YOLOv8 + SAM framework reveals that approach achieves scores of up to 0.97 of IoU, representing an improvement of approximately 8–10 percentage points over YOLOv8 alone across different scenarios. Notably, a minor decrease is observed in shadowed conditions for PTC and Fresnel mirrors, primarily due to YOLO's detection challenges. The proposed framework provides a promising foundation for developing AI-driven, automated O&M solutions that enhance real-time maintenance in CSP systems.