摘要
将冷轧带钢表面缺陷图像中的所有像素作为高维空间中的特征向量,利用有监督非线性降维方法对其进行减维后再进行缺陷的分类。该方法解决了冷轧带钢表面缺陷自动分类中的特征提取和特征选择的困难,避免了分类器特征维数过高的问题,并可以用于动态数据的在线识别和聚类。用这种降维方法并结合K近邻分类器与支持向量机对现场采集到的缺陷样本数据集进行试验,结果表明经过降维预处理后,2种分类器的性能都得到了很大的提高。
All pixels of surface defect images of cold rolled strips are regarded as feature vectors in a high dimensional space, and then they are preprocessed by supervised nonlinear dimensionality reduction before classification. The method can ,solve the problems of feature extraction and selection, avoid the difficulties caused by huge dimensional data, and can be used in online clustering and recognizing dynamic data. Experiment on industrial scale was done by combining the method with K nearest neighbor classifier and support vector machine. The results show that performance of two kinds of classifiers is enhanced greatly after dimensionality reduction preprocessing.
出处
《钢铁》
CAS
CSCD
北大核心
2005年第12期37-40,共4页
Iron and Steel
基金
国家自然科学基金资助项目(50074010)
国家高技术研究发展(863)计划资助项目(2001AA339030)
关键词
非线性降维
冷轧带钢
表面缺陷
分类
nonlinear dimensionality reduction
cold rolled strip
surface defect
classification
