东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2007, Vol. 28 ›› Issue (3): 410-413.DOI: -

• 论著 • 上一篇    下一篇

基于改进支持向量机的冷轧带钢表面缺陷分类识别

王成明;颜云辉;陈世礼;韩英莉;   

  1. 东北大学机械工程与自动化学院;东北大学机械工程与自动化学院;东软集团有限公司商用软件事业部;东北大学机械工程与自动化学院 辽宁沈阳110004;辽宁沈阳110004;辽宁沈阳110179;辽宁沈阳110004
  • 收稿日期:2013-06-24 修回日期:2013-06-24 出版日期:2007-03-15 发布日期:2013-06-24
  • 通讯作者: Wang, C.-M.
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(50574019);;

Classification and recognition based on improved SVM for surface defects of cold strips

Wang, Cheng-Ming (1); Yan, Yun-Hui (1); Chen, Shi-Li (2); Han, Ying-Li (1)   

  1. (1) School of Mechanical Engineering and Automation, Northeastern University, Shenyang 110004, China; (2) Business Software Department, Neusoft Group Ltd., Shenyang 110179, China
  • Received:2013-06-24 Revised:2013-06-24 Online:2007-03-15 Published:2013-06-24
  • Contact: Wang, C.-M.
  • About author:-
  • Supported by:
    -

RichHTML

1

PDF (PC)

587

摘要： 针对冷轧带钢表面缺陷图像模式识别中存在的问题,提出了基于改进支持向量机的冷轧带钢典型表面缺陷分类识别方法.根据带钢表面缺陷图像识别的特点,对渐进直推式支持向量机在其基本原理基础上加以改进,设计了一种冷轧带钢表面缺陷图像模式识别的分类器.通过实验确定了分类器的结构,给出了相关参数选择的方法.对几种生产现场出现频率较高的典型缺陷图像进行了计算机实验研究.研究结果显示,这种分类器很好地克服了传统支持向量机中存在的推广性能差以及当类别距离过近时准确率下降的问题,具有更好的适应性和准确性.

关键词: 冷轧带钢, 表面缺陷, 分类识别, 支持向量机, 分类器

Abstract: Aiming at the existing problems in pattern recognition of surface defect images of cold strips, a classification and recognition method is proposed to solve them, based on improved support vector machine (SVM). According to the features in recognizing those surface defect images and on the basic principle of SVM, the method provides an effective improvement to the progressively immediate inference SVM, thus designing a pattern recognition classifier for those surface defect images. The structure of the classifier is confirmed by experiments, with a method given to choose relevant parameters. Experimental investigation was carried out on computer aiming at several typical defect images which were found frequently in site, and the results showed that the classifier based on the improved SVM can solve the problems the conventional SVM is unable to solve, such as poor generality and decreasing accuracy when the gap between two classes is too narrow. In addition, the classifier provides higher adaptability and accuracy.

中图分类号: