皮革科学与工程 ›› 2022, Vol. 32 ›› Issue (4): 71-76.doi: 10.19677/j.issn.1004-7964.2022.04.012

• 标准化与检验 • 上一篇    下一篇

红外光谱联合主成分分析在貂毛皮鉴别中的应用

袁绪政1,2,*, 李鹏妮3, 张红1, 庄莉2   

  1. 1.嘉兴市皮毛和制鞋工业研究所,浙江 桐乡 314500;
    2.国家纺织服装产品质量检验中心浙江 桐乡,浙江 桐乡 314500;
    3.桐乡市质量技术监督事务中心,浙江 桐乡 314500
  • 收稿日期:2022-01-05 出版日期:2022-08-28 上线日期:2022-07-11
  • 通讯作者: *
  • 作者简介:袁绪政(1983-),男,博士,高级工程师,主要从事皮革、毛皮检验检测工作,E-mail:175745678@qq.com。
  • 基金资助:
    浙江省市场监督管理局科研计划项目(20190340)

Application of Infrared Spectroscopy Combined with Principal Component Analysis in the Identification of Mink Fur

YUAN Xuzheng1,2,*, LI Pengni3, ZHANG Hong1, ZHUANG Li2   

  1. 1. Jiaxing Fur and Footwear Research Institute, Tongxiang 314500, China;
    2. National Wool Knitwear Quality Supervision Inspection Center, Tongxiang 314500, China;
    3. Tongxiang Affairs Center of Quality and Technical Supervision, Tongxiang 314500, China
  • Received:2022-01-05 Online:2022-08-28 Published:2022-07-11

摘要: 基于红外光谱,联合SPSS软件的主成分分析和分类判别方法对60组貂毛皮与60组易混非貂毛皮进行鉴别分析(随机抽取100组数据作为训练数据,余下20组作为验证数据)。结果表明:通过主成分分析可以从大量光谱信息中提取有用的信息,成功地将原2696个波数变量降维到8个新的主成分变量,使问题得到简化,提高了分析效率;通过分类判别成功建立了貂毛皮与非貂毛皮的典型判别函数、貂毛皮和非貂毛皮的Bayes分类函数。多种验证方法表明:100组训练数据的典型判别函数、Bayes分类函数自身验证分类正确率均为97.0%、交叉验证分类正确率为95.0%;20组验证数据分类正确率达100%。红外光谱联合主成分分析用于貂毛皮真假鉴别效果良好。

关键词: 红外光谱, 主成分分析, 分类判别, 材质鉴别, 貂毛皮

Abstract: Using the infrared spectroscopy method, combined with the SPSS (Statistical Product and Service Solutions) software method of principal component analysis and discriminant analysis, 60 groups of mink fur samples and 60 groups of easily mixed non-mink fur samples were identified, in which 100 groups data were randomly selected as the training data while the remaining 20 groups as the verification data. Results show that the principal component analysis could extract useful information from a large amount of spectral information. The dimension of the original 2696 wavenumber variables was successfully reduced to 8 new principal component variables. As a result, the problem was simplified and the analysis efficiency was improved. The typical discriminant functions and the Bayes classification functions of mink and non-mink fur were obtained by classification discriminant analysis. Results from a variety of verification methods show that the self-verification classification accuracies of both the typical discriminant function and Bayes classification function of 100 groups of training data are 97.00%, while the cross-verification classification accuracy is 95.00% and the classification accuracy of 20 groups of verification data is 100%. In short, the infrared spectroscopy method combined with the principal component analysis is effective in the identification of mink fur.

Key words: infrared spectroscopy, principal component analysis, discriminant analysis, material identification, mink fur

中图分类号: