光谱学与光谱分析, 2019, 39 (5): 1653, 网络出版: 2019-05-13  

基于二维相关光谱分析的纯棉与丝光棉制品鉴别分析

Identification of Pure Cotton and Mercerized Cotton Fabrics Based on 2-D Correlation Spectra Analysis
作者单位
1 北京化工大学信息科学与技术学院, 北京 100029
2 北京化工大学材料科学与技术学院, 北京 100029
摘要
纯棉与丝光棉制品是日常生活中常用的两种纤维制品, 但是由于二者在物理结构和化学结构上非常相似, 以至于使用一些简单的方法难以准确识别一部分纯棉与丝光棉制品。 提出一种使用水含量作为扰动的二维相关光谱结合机器学习方法来对二者进行鉴别的新方法。 共使用从专业机构获得的200个标准样本来设计实验对新方法进行验证, 其中包括100个纯棉样本与100个丝光棉样本。 对每一个样本, 使用水含量作为扰动, 分4次改变样本水含量并采集该水含量下样本的一维光谱, 其中4次的水含量分别为20.20%, 14.52%, 7.77%与0%。 根据四条不同的一维构造每一个样本的动态光谱, 再通过二维相关算法来计算其同步二维相关光谱, 从该同步二维相关光谱中使用移动窗口技术提取三组不同的分类特征, 每组特征分别对应一个设计好的支持向量机(SVM)分类器。 之后本文提出一种基于信息熵的多分类器融合方法, 根据权值不同, 将三个分类器融合为一个具有更优效果的强分类器。 为了验证方法的准确性与有效性, 设计了严谨的实验对方法进行验证。 实验首先按照传统的从一维光谱中提取特征的方法对纯棉与丝光棉样本进行鉴别, 使用两种样本各50个来进行分类模型建立, 剩余的进行模型验证, 分类效果最高只有76%。 但是基于从二维相关光谱中提取的三组特征设计的三个支持向量机(SVM)分类器的准确率分别可以达到88%, 90%, 88%, 最后根据提出的基于信息熵的多分类器信息融合方法将三个分类器进行融合同一可以得到92%的分类准确率, 比三个基础分类器准确率都有提升。 与从一维光谱中提取特征并设计分类器进行分别鉴别相比, 从二维相关光谱中提取特征设计多个分类器并使用基于信息熵的多分类器信息融合方法进行分类鉴别具有更高的分类准确率。 二维相关光谱将光谱信息扩展到更高的维度, 将一维光谱中隐藏的折叠峰进行展开, 因此具有更高的分类准确率。 提出的方法是一种快速准确鉴别纯棉与丝光棉制品的新方法。
Abstract
Pure cotton and mercerized cotton products are widely used in daily life. It is difficult to classify the pure cotton and mercerized cotton products with simple methods because they are similar in chemical and physical structures. In this work, a new method of rapid identification of pure cotton and mercerized cotton products with two-dimension correlation spectra analysis was proposed. In this work, 200 textile samples including 100 pure cotton fiber products and 100 mercerized cotton fiber products were collected. For each sample, the water content was changed 4 times and one-dimension spectra was collected, among them, the water content of 4 times was 20.20%, 14.52%, 7.77% and 0% respectively. Then their simultaneous two-dimension correlation spectra were calculated based on correlational analysis. Three kinds of classification features were extracted from the synchronous two-dimension correlation spectra. Support Vector Machine (SVM) was combined with different kind of the classification features to construct different classifiers. In this work, an information fusion method was proposed to make the multi-classifier decision. To verify the feasibility and effectiveness of the proposed method, the comparative experiments have been done. The accuracy of identification with the classifier based on extracted one-dimensional spectra features with PCA was only 76%. The accuracy of identification with the three classifiers based on extracted features from two-dimensional correlation spectra were 88%, 90% and 88% respectively. The accuracy of identification with the proposed method was 92%. Compared with one-dimension spectra based feature extraction, the two-dimension correlation spectra based feature extraction achieved feature enhancement and the multi-classifier fusion decision could improve the accuracy of classification obviously. Two-dimensional correlation spectroscopy extended spectral information to higher dimensions, unfolded hidden fold peaks in one-dimensional spectra, and had higher classification accuracy. The proposed method provided a new way for rapid identification of pure cotton and mercerized cotton products.

曹凯, 赵众, 袁洪福, 李彬. 基于二维相关光谱分析的纯棉与丝光棉制品鉴别分析[J]. 光谱学与光谱分析, 2019, 39(5): 1653. CAO Kai, ZHAO Zhong, YUAN Hong-fu, LI Bin. Identification of Pure Cotton and Mercerized Cotton Fabrics Based on 2-D Correlation Spectra Analysis[J]. Spectroscopy and Spectral Analysis, 2019, 39(5): 1653.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!