多层次分割实现微小尺寸实时测量

第32卷第2期光电工程Vol.32, No.2 2005年2月 Opto-Electronic Engineering Feb, 2005

文章编号

崔雁松３　

２．江苏工业学院　３．天津职业大学　摘要提出了两次阈值分割的检测算法

而后分割是在过渡区搜索目标真实边缘实验表明精度优于0.001mm

?úêμê±D??üé?比拟合算法缩短了40%

关键词图像处理微尺寸测量

中图分类号A

Micro-dimension detection by means of multi-level segmentation

CHEN Yu1, ZHANG Xin-rong2, ZHU Yi-wei3, CUI Yan-song4

(1. College of Electronic Information Engineering, Tianjin University, Tianjin 300072, China;

2. Jiangsu Polytechnical University, Changzhou 213016, China;

3. Tianjin Professional College, Tianjin 300171, China)

Abstract: For the purpose of real-time micro-dimension detection of the multiple ramp-edged objects in

a grey scale image, a twice-segmentation detection algorithm is proposed. The first segmentation is to

extract object and determine transition area while the second segmentation is to search the real edges of an object in the transition area for detecting micro-dimension of the object. Experiments show that this twice threshold segmentation algorithm has obvious advantages for measuring objects smaller than

0.5mm with better measuring accuracy than 0.001mm and the influence of noise on real edge point

number is reduced. The average measuring time for an object is 60ms which is only 60% of that of fitting algorithm. This guarantees 2000 packages/h processing rate of the detection system.

Key words: Vision detection; Threshold segmentation; Image processing; Micro-dimension detection

引 言　

微小尺寸图像检测作为非接触尺寸检测技术的分支

从图像本身看而微尺寸检测的本质就是精确提取目标边缘

即目标实际获得的灰度图像噪声明显

这给边缘的提取带来困难

如 Robert cross 算子Sobel算子进行对于上述条件的灰度图像

二阶算子产生过多伪边缘点

比起并行边界分割技术由于噪声的

收稿日期收到修改稿日期

燏汉族江苏无锡人主要从事图像处理和模式识别