基于稀疏字典学习的羊绒与羊毛分类  

Cashmere and wool classification based on sparse dictionary learning

作　　者：孙春红 丁广太[1,2] 方坤 SUN Chunhong;DING Guangtai;FANG Kun(School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China;Materials Genome Institute, Shanghai University, Shanghai 200444, China)

机构地区：[1]上海大学计算机工程与科学学院,上海200444 [2]上海大学材料基因组工程研究院,上海200444

出　　处：《纺织学报》2022年第4期28-32,39,共6页Journal of Textile Research

基　　金：国家重点研发计划项目(2018YFB0704400,2016YFB0700500)。

摘　　要：为准确鉴别羊绒与羊毛纤维,提出了一种基于稀疏字典学习的分类方法。首先,对纤维图像进行预处理实现数据增强,获取纤维图像特征矩阵;之后,对特征矩阵进行字典学习,获取过完备字典与稀疏编码;最后,通过稀疏编码与字典实现羊绒与羊毛的分类和鉴别。该方法使用光学显微镜以及扫描电子显微镜图像作为数据集,实验结果表明,与支持向量机分类器以及基于稀疏表示的分类算法相比,本文方法的分类准确率可提高5%~10%,分类准确率最高可达到91%,可用于后续实际的羊绒与羊毛纤维分类与鉴定工作。In order to identify cashmere and wool fibers accurately,this paper proposes a classification method based on sparse dictionary learning.Firstly,the fiber image is preprocessed to achieve data enhancement to achieve a fiber image feature matrix.Secondly,dictionary learning is performed on the feature matrix to obtain a complete dictionary and sparse coding.Finally,based on sparse coding and dictionary,the classification and identification of cashmere and wool is implemented.This method uses optical microscope images and scanning electron microscope images as data sets.Experiment results show that compared with support vector machine classifiers and sparse representation-based classifier algorithms,the classification accuracy of this method can be improved by 5%-10%,and the classification accuracy can reach up to 91%.It can be used for subsequent actual classification and identification of cashmere and wool fibers.

关 键 词：稀疏表示 字典学习 图像识别 机器学习 羊绒 羊毛 纤维鉴别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP131.9[自动化与计算机技术—计算机科学与技术]