摘要
将LVQ神经网络用于冷轧带钢表面缺陷的自动分类中，解决了以往分类方法在多缺陷模式类型情况下耗时多和准确率低的问题．对现场采集到的14种主要缺陷类型进行了实验．实验结果表明，基于LVQ神经网络的分类器训练与分类的时间短，在多缺陷种类分类的过程中准确率能得到保证．
A new method which uses LVQ neural network in the automatic classification of surface defects for cold rolled strips was presented. The problems of long time and low accuracy in the classification of multi-defect pattern types with some traditional classification algorithms were resolved. Tested by 14 main defect types collected from online data, the results demonstrated that the method of surface defects for cold rolled strips based on LVQ neural network spent little time during training and classifying, and its accuracy could be assured on the recognition process of multi-defect pattern types.
出处
《北京科技大学学报》
EI
CAS
CSCD
北大核心
2005年第6期732-735,共4页
Journal of University of Science and Technology Beijing
基金
国家自然科学基金(No.50074010)国家"863"计划课题(No.2003AA331080)
关键词
冷轧带钢
表面缺陷
缺陷分类
LVQ神经网络
cold rolled strips
surface defect
defect classification
LVQ neural network
