零件分拣系统中图像处理算法研究

作者：单晓杭，侯海鹏，张利 日期：2010-06-28/span> 浏览：3358 查看PDF文档

零件分拣系统中图像处理算法研究
单晓杭，侯海鹏，张利
(浙江工业大学 特种装备制造与先进加工技术教育部重点实验室，浙江 杭州 310032)

摘要：为了提高生产中机械零件自动识别的实时性及分拣的准确率，提出了一种二次分割算法，对链片上有无钢印进行识别研究。首先，对采集的图像进行了基于Otsu法的背景自动阈值分割；接着，通过质心定位提取质心邻域矩阵，实现了特征区域图像的分割；最后，给出了基于特征区灰度值统计模型的判别式，并对关键参数进行了优化。试验结果表明所提出的二次分割算法准确率高、鲁棒性好，能有效满足复杂环境下特征对象的识别要求。
关键词：目标识别；自动阈值；图像分割；质心定位
中图分类号：TP391.4文献标识码：A文章编号：1001-4551(2010)05-0094-04

Research of image processing algorithm for parts sorting system

SHAN Xiaohang, HOU Haipeng, ZHANG Li
(The MOE Key Laboratory of Special Purpose Equipment and Advanced Processing,
Zhejiang University of Technology, Hangzhou 310032, China)

Abstract: Recognition of steel seal on mechanical chain piece by image processing was investigated in order to classify better in produce, and a twice segmentation method was proposed according to the analysis of parts characters. Firstly, automatic thresholding using the Otsu algorithm for image sampled was performed. And a neighborhood matrix was extracted by centroid localization, so that a preferable segmentation of character region concerned was gained. At last, a discriminant based on grayscale statistical model was given. Moreover, some key parameters were optimized. The research results show that the algorithm shows high accuracy and good robustness.
Key words: object recognition; automatic thresholding; image segmentation; centroid localization