华南理工大学学报(自然科学版) ›› 2003, Vol. 31 ›› Issue (12): 30-33.

• • 上一篇    下一篇

机械零件图像中直线边缘亚像素定位方法

杨 敏, 叶邦彦, 牟 丽, 吴春凌    

  1.  华南理工大学 机械工程学院‚广东 广州510640
  • 出版日期:2003-12-20 发布日期:2022-09-01
  • 作者简介:杨敏(1974-)‚男‚博士生‚主要从事计算机图 像处理、机器视觉和模式识别等研究.

Method of Straight Line Edge Subpixel Localization for Mechanical Part Image

Yang Min, Ye Bang-yan, Mou Li, Wu Chun-ling   

  1. College of Mechanical Eng.‚South China Univ.of Tech.‚Guangzhou510640‚China
  • Online:2003-12-20 Published:2022-09-01

摘要: 以计算机视觉在机械零件几何参数检测的实际应用为例‚提出了一种使用高斯 拉普拉斯(LOG)函数检测斜坡状边缘的像素级位置‚在已知目标为直线边缘的情况下‚ 使用最小二乘线性回归把二维的边缘拟合降为一维边缘定位‚从而使直线边缘定位达到 亚像素级精度的算法.同时还进行了空间矩直线边缘亚像素定位算法与本文所提出算法 的对比实验.实验结果表明:在低噪声图像中‚两种算法的边缘定位精度均达到满意的结 果‚且最小二乘线性回归亚像素定位算法速度较快.

关键词: 边缘检测, LOG 算子, 线性回归, 空间矩, 亚像素

Abstract: Based on practical application of computer vision for mechanical parts geometry parameter inspection‚a new approach is presented to determine the ramp edge at the level of pixel using Laplacian of Gaussian (LOG) function.The pixel level edge to subpixel accuracy is then located by using the MLS linear regression on the condition that the model of the edge is known.Contrastive experiments have been conducted to test the algorithm of the spatial moment subpixel localization for the straight line edge and the algorithm proposed in this paper.The experiment results show that the precision of two algorithms is satisfactory and the velocity of the MLS linear regression subpixel localization is faster. 

Key words: edge inspection, LOG operator, linear regression, spatial moment, subpixel

中图分类号: