Scholarlink

<p>In the context of intelligent manufacturing, a method for measuring defects in cemented carbide circular saw blades based on machine vision is proposed to eliminate the need for prior knowledge during manual inspection and the low detection efficiency during the production and operation of circular saw blades. This method enables rapid defect measurement through a combination of deep learning and traditional image processing. First, the improved You Only Look Once v4-tiny model extracts the target sawtooth and eliminates complex background interference. Second, the Otsu method is integrated with the Canny operator to extract the sawtooth edge, and subpixel extraction is applied to enhance edge accuracy. Third, the edge points to be inspected are registered with standard edge points, utilizing a contour point distribution histogram for coarse registration, followed by fine registration using the improved iterative closest point. Finally, dynamic time warping is applied to measure and evaluate the registered sawtooth edge. Experimental results show that this method achieves a mean average precision value of 99.99% in detecting detached sawtooth. The average relative measurement errors for normal, worn, and crushed sawteeth are approximately 11.2%, 13.0%, and 5.9%, respectively. This method effectively measures small defects on the sawtooth and provides a new approach to measuring tool defects.</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

Defect measurement method of circular saw blade based on machine vision

  • 期刊论文

  • 发表日期：

    2025年2月19日

查看全文
  • Hui Wang,
  • Yangyu Wang,
  • Pengcheng Ni,
  • Gonghua Lan,
  • Deguang Liu,
  • Guojian He,
  • Weiguang Lou,
  • Erzhong Feng

摘要

In the context of intelligent manufacturing, a method for measuring defects in cemented carbide circular saw blades based on machine vision is proposed to eliminate the need for prior knowledge during manual inspection and the low detection efficiency during the production and operation of circular saw blades. This method enables rapid defect measurement through a combination of deep learning and traditional image processing. First, the improved You Only Look Once v4-tiny model extracts the target sawtooth and eliminates complex background interference. Second, the Otsu method is integrated with the Canny operator to extract the sawtooth edge, and subpixel extraction is applied to enhance edge accuracy. Third, the edge points to be inspected are registered with standard edge points, utilizing a contour point distribution histogram for coarse registration, followed by fine registration using the improved iterative closest point. Finally, dynamic time warping is applied to measure and evaluate the registered sawtooth edge. Experimental results show that this method achieves a mean average precision value of 99.99% in detecting detached sawtooth. The average relative measurement errors for normal, worn, and crushed sawteeth are approximately 11.2%, 13.0%, and 5.9%, respectively. This method effectively measures small defects on the sawtooth and provides a new approach to measuring tool defects.