Scholarlink

<p>Surface defect inspection constitutes a pivotal quality control phase in modern manufacturing processes, serving as a crucial safeguard against market circulation of defective products. This review presents a systematic examination of cutting-edge methodologies by synthesizing recent advancements in empirical mode decomposition (EMD) and artificial intelligence technologies. The analysis begins with an in-depth exploration of EMD’s unique advantages as an adaptive signal processing mechanism, particularly its exceptional capability in addressing nonlinearity and non-stationarity characteristics inherent in product surface defect signals. Subsequently, the paper conducts a comprehensive evaluation of deep learning applications through five methodological frameworks: self-supervised, semi-supervised, fully supervised, weakly supervised, and unsupervised learning paradigms, highlighting their respective strengths in industrial visual inspection scenarios. The study further systematically categorizes current challenges in surface defect detection across different manufacturing environments while presenting targeted solutions supported by experimental validations. Finally, the integration of EMD with deep learning is explored, emphasizing its potential and technological significance for enhancing surface defect detection capabilities. This work offers valuable insights and guidance for future research and development in this critical field.</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

Research progress on key technologies for product surface defect detection based on empirical mode decomposition and deep learning

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Jianbin Xiong,
  • Qianguang Zhang,
  • Qi Wang,
  • Jianxiang Yang,
  • Xiangjun Dong,
  • Qiong Liang

摘要

Surface defect inspection constitutes a pivotal quality control phase in modern manufacturing processes, serving as a crucial safeguard against market circulation of defective products. This review presents a systematic examination of cutting-edge methodologies by synthesizing recent advancements in empirical mode decomposition (EMD) and artificial intelligence technologies. The analysis begins with an in-depth exploration of EMD’s unique advantages as an adaptive signal processing mechanism, particularly its exceptional capability in addressing nonlinearity and non-stationarity characteristics inherent in product surface defect signals. Subsequently, the paper conducts a comprehensive evaluation of deep learning applications through five methodological frameworks: self-supervised, semi-supervised, fully supervised, weakly supervised, and unsupervised learning paradigms, highlighting their respective strengths in industrial visual inspection scenarios. The study further systematically categorizes current challenges in surface defect detection across different manufacturing environments while presenting targeted solutions supported by experimental validations. Finally, the integration of EMD with deep learning is explored, emphasizing its potential and technological significance for enhancing surface defect detection capabilities. This work offers valuable insights and guidance for future research and development in this critical field.