Scholarlink

<p>Ellipse detection is a critical task in computer vision with applications ranging from medical imaging to traffic sign recognition. Despite significant advancements, achieving accurate ellipse detection under challenging conditions remains a formidable challenge. This paper introduces BEANet, a novel boundary-enhanced attention network designed specifically for precise ellipse detection. BEANet incorporates a boundary-enhanced attention module that emphasizes informative boundary features, thereby improving the accuracy of ellipse regression. To address boundary discontinuity issues in ellipse angle regression, we propose a periodic angular loss and utilize the Gaussian Wasserstein Distance loss to enhance geometric stability. Experiments on the public GED dataset demonstrate that BEANet outperforms existing state-of-the-art methods, achieving up to 9.94% higher average precision. Our results highlight the effectiveness of our approach in detecting ellipses under complex conditions, paving the way for more robust and accurate ellipse detection in various applications.</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

Enhancing ellipse detection precision via boundary-aware attention and geometric loss optimization

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Ouxiang Chen,
  • Hongcui Wang,
  • Lin Li

摘要

Ellipse detection is a critical task in computer vision with applications ranging from medical imaging to traffic sign recognition. Despite significant advancements, achieving accurate ellipse detection under challenging conditions remains a formidable challenge. This paper introduces BEANet, a novel boundary-enhanced attention network designed specifically for precise ellipse detection. BEANet incorporates a boundary-enhanced attention module that emphasizes informative boundary features, thereby improving the accuracy of ellipse regression. To address boundary discontinuity issues in ellipse angle regression, we propose a periodic angular loss and utilize the Gaussian Wasserstein Distance loss to enhance geometric stability. Experiments on the public GED dataset demonstrate that BEANet outperforms existing state-of-the-art methods, achieving up to 9.94% higher average precision. Our results highlight the effectiveness of our approach in detecting ellipses under complex conditions, paving the way for more robust and accurate ellipse detection in various applications.