Scholarlink

<p>Autonomous inspection robots operating in complex transformer oil environments are essential for real-time monitoring. However, their limited field of view compromises visual perception and user experience. To address this issue, we propose an efficient and accurate image stitching method that integrates Xfeat-based feature extraction, dual homography matrix alignment, and diffusion model-driven rectification. Xfeat enhances feature detection and matching, ensuring robustness in complex transformer oil environments. The dual homography matrix method improves image alignment by adapting to spatial variations, thereby reducing distortion in stitched images. Finally, by leveraging diffusion models, we refine image boundaries, mitigate ghosting effects, and ensure seamless integration. Experimental results demonstrate that our approach outperforms conventional stitching techniques, producing clearer and more stable images with higher efficiency, even in complex transformer oil conditions. The proposed framework significantly enhances the visual interaction experience of inspection robots, facilitating more reliable condition monitoring of transformers.</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

Accurate and efficient image stitching for inspection robots working in complex transformer oil environments

  • 期刊论文

  • 发表日期：

    2025年4月15日

查看全文
  • Liqing Liu,
  • Chun He,
  • Chuang Yao,
  • Junji Feng,
  • Cheng He,
  • Haochong Chen,
  • Xuebo Zhang

摘要

Autonomous inspection robots operating in complex transformer oil environments are essential for real-time monitoring. However, their limited field of view compromises visual perception and user experience. To address this issue, we propose an efficient and accurate image stitching method that integrates Xfeat-based feature extraction, dual homography matrix alignment, and diffusion model-driven rectification. Xfeat enhances feature detection and matching, ensuring robustness in complex transformer oil environments. The dual homography matrix method improves image alignment by adapting to spatial variations, thereby reducing distortion in stitched images. Finally, by leveraging diffusion models, we refine image boundaries, mitigate ghosting effects, and ensure seamless integration. Experimental results demonstrate that our approach outperforms conventional stitching techniques, producing clearer and more stable images with higher efficiency, even in complex transformer oil conditions. The proposed framework significantly enhances the visual interaction experience of inspection robots, facilitating more reliable condition monitoring of transformers.