Scholarlink

<p>The weld pass classification and the region of interest (ROI) extraction are crucial for enhancing the subsequent feature extraction and image processing speed in robotic automatic multi-layer multi-pass (MLMP) welding. Therefore, a novel model, lightweight weld pass classification and ROI extraction (LCE), is proposed. Initially, LCE is constructed based on object detection by some lightweight modules and structures, achieving a low model scale. Then, an enhanced bounding box regression loss function is presented to improve the precision of LCE. Finally, experiments show that LCE’s mean average precision and processing time are 92.7% and 7.2&#xa0;ms, respectively. Furthermore, the software encapsulated with LCE also achieves high precision and speed. The main novelties and contributions of this paper include a novel lightweight model and an enhanced loss function, advancing the field of industrial automatic welding.</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

Lightweight weld pass classification and region of interest extraction for robotic multi-layer multi-pass welding

  • 期刊论文

  • 发表日期：

    2026年3月16日

查看全文
  • Hui Xu,
  • Yingjie Guo,
  • Huiyue Dong,
  • Minghua Zhu,
  • Xiaofan Dai,
  • Zhi Li,
  • Yinglin Ke

摘要

The weld pass classification and the region of interest (ROI) extraction are crucial for enhancing the subsequent feature extraction and image processing speed in robotic automatic multi-layer multi-pass (MLMP) welding. Therefore, a novel model, lightweight weld pass classification and ROI extraction (LCE), is proposed. Initially, LCE is constructed based on object detection by some lightweight modules and structures, achieving a low model scale. Then, an enhanced bounding box regression loss function is presented to improve the precision of LCE. Finally, experiments show that LCE’s mean average precision and processing time are 92.7% and 7.2 ms, respectively. Furthermore, the software encapsulated with LCE also achieves high precision and speed. The main novelties and contributions of this paper include a novel lightweight model and an enhanced loss function, advancing the field of industrial automatic welding.