华南理工大学学报(自然科学版) ›› 2006, Vol. 34 ›› Issue (11): 55-59.

• 机械工程 • 上一篇    下一篇

基于计算机视觉技术的微钻刃面自动光学检测

张舞杰 李迪 叶峰   

  1. 华南理丁大学 机械工程学院,广东 广州 510640
  • 出版日期:2006-11-25 发布日期:2006-11-25
  • 通信作者: 张舞杰(1970-),男,博士生,主要从事图像处理、模式识别、过程监控、嵌入式装备控制等方面的研究 E-mail:zhang_ liu73@163.corn
  • 作者简介:张舞杰(1970-),男,博士生,主要从事图像处理、模式识别、过程监控、嵌入式装备控制等方面的研究.
  • 基金资助:

    粤港关键领域重点突破项目(2004A10403021)

Automatic Optical Inspection of M ini-Drill Blade Based on Computer Vision Technology

Zhang Wu-jie  Li Di  Ye Feng   

  1. School of Mechanical Engineering,South China Univ.ofTech.,Guangzhou 510640,Guangdong,China
  • Online:2006-11-25 Published:2006-11-25
  • Contact: 张舞杰(1970-),男,博士生,主要从事图像处理、模式识别、过程监控、嵌入式装备控制等方面的研究 E-mail:zhang_ liu73@163.corn
  • About author:张舞杰(1970-),男,博士生,主要从事图像处理、模式识别、过程监控、嵌入式装备控制等方面的研究
  • Supported by:

    粤港关键领域重点突破项目(2004A10403021)

摘要: 为了实现微钻刃面尺寸和缺陷的自动光学检测,建立了影像检测系统并研究其图像处理算法.根据微钻刃面的特征,设计了专用的照明系统来获取清晰的、变形小的微钻刃面图像.采用改进的小核值相似区边缘提取算法对图像进行边缘提取,并对所提取的边缘采用基于空间矩的亚像素算法进行图像边缘的亚像素定位,然后采用直线和圆弧拟合等一系列算法对微钻刃面图像进行尺寸计算和缺陷检测.实验结果表明:自动光学检测方法的准确率达到了99.5%,微钻的检出精度达到了微米级,说明该方法能够满足微钻刃面尺寸测量和缺陷检测的要求.

关键词: 计算机视觉, 自动光学测量, 边缘检测, 亚像素, 小核值相似区算法

Abstract:

In order to realize the automatic optical inspection(AOI)of the dimensions and defects of mini.drill blades,an automatic image inspection system is proposed and the corresponding image processing algorithms are in-vestigated.In the investigation,a special lighting system is first designed based on the characteristics of mini-drill blades to obtain clear blade images with little distortion.Next,the improved Small Univalue Segment Assimilating Nucleus edge extraction algorithm is adopted to extract the image edge.Then,the extracted edge is located by the subpixel edge location algorithm based on the spatial moments to obtain a precise Subpixel location.Finally,the algo-rithms,such as the line fitting and the circle fitting,are adopted to calculate the dimensions and to inspect the de-fects of the mini-drill blades.Experimental results demonstrate that the proposed AOI method is of an inspection ac-curacy of 99.5% and a dimension precision in the scale of micron.so it can meet the requirements of dimension measurement and defect inspection of mini-drill blades.

Key words: computer vision, automatic optical inspection, edge detection, subpixel, Small Univalue Segment As-similating Nucleus Aalgorithm