Tao Shuai, Ma Xiang, Li Jun-Hui, Zhang Wen-Juan, Zhao Long-Lian, Wen Ya-Dong, Wang Yi, Zhang Lu-Da
College of Information and Electrical Engineering, China Agricultural University, Beijing 100193, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2009 Nov;29(11):2970-4.
The present paper briefly describes the application of near-infrared spectroscopy technology in tobacco. Two methods for solving projection vector based on the principal component of near-infrared spectrum are reported. They are named as projection of basing on principal component and Fisher criterion by principal component analysis method (PPF_PCA for short) and projection of basing on principal component and Fisher criterion by solving eigenvalue (PPF_Eig for short), and they are studied and compared in the application and evaluation of the redrying model. The result of the first-dimensional projection on 9 kinds of tobacco leaf grading samples shows that, the diversity of the first-dimensional projection values of inter-class and intra-class by the PPF_PCA method is both larger than that by the PPF_Eig method, and the mean absolute deviation of the mean projection values of inter-class by the PPF_PCA method is about 1.26 times that of the PPF_Eig method. At the same time, this result is interpreted by using the contribution rate of the first-dimensional projection values. That is, the contribution rate of first-dimensional projection values by the PPF_PCA method is 93%, while the contribution rate of first-dimensional projection values by the PPF_Eig method is 77%. The former is about 1.21 times the later. Therefore, the first-dimensional projection values by PPF PCA method include more information of diversity of both inter-class and intra-class. The similarity of samples inter-class and the diversity of samples intra-class can be evaluated more objectively from first-dimensional projection figure(on 9 kinds of tobacco leaf grading samples, 33 kinds of tobacco leaf grading samples and 6 redrying blending models), so it is more convenient to be used as a reference for the redrying model of tobacco, and it has a good application prospect in other formulation design of agricultural products (traditional Chinese medicine etc.).
本文简要介绍了近红外光谱技术在烟草中的应用。报道了两种基于近红外光谱主成分求解投影向量的方法。它们分别是主成分分析法基于主成分和Fisher准则的投影(简称PPF_PCA)和求解特征值法基于主成分和Fisher准则的投影(简称PPF_Eig),并在复烤模型的应用和评价中对它们进行了研究和比较。对9种烟叶分级样品的一维投影结果表明,PPF_PCA法的类间和类内一维投影值的差异均大于PPF_Eig法,PPF_PCA法类间平均投影值的平均绝对偏差约为PPF_Eig法的1.26倍。同时,利用一维投影值的贡献率对该结果进行了解释。即PPF_PCA法的一维投影值贡献率为93%,而PPF_Eig法的一维投影值贡献率为77%。前者约为后者的1.21倍。因此,PPF_PCA法的一维投影值包含了更多的类间和类内差异信息。从一维投影图(对9种烟叶分级样品、33种烟叶分级样品和6种复烤配方模型)上可以更客观地评价样品的类间相似度和类内差异度,因此更便于作为烟草复烤模型的参考,在其他农产品(如中药等)配方设计中也具有良好的应用前景。