标题：Comparison of Different Spectral Preprocessing Methods Onidentification of Green Tea Origins by Near Infrared Spectroscopy
作者：Zhuang, X. G.; Wang, L. L.; Fang, J. X.; Wu, X. Y.
作者机构：[Zhuang, X. G.; Wang, L. L.; Fang, J. X.] Shandong Univ, Adv Res Ctr Opt, Jinan 250100, Peoples R China.; [Zhuang, X. G.; Wang, L. L.; Fang, J. X.] 更多
会议名称：International Conference on Computer Science and Environmental Engineering (CSEE)
会议日期：MAY 17-18, 2015
来源：INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND ENVIRONMENTAL ENGINEERING (CSEE 2015)
关键词：NIR spectroscopy; SVM; spectral preprocessing; green tea
摘要：In this study, several common spectral preprocessing methodsin near infrared (NIR) spectroscopy for rapid and non-destructive identification ofgreen tea origins were given systematicallyinvestigated. All spectra of 60green tea samples were collectedby a low-resolution spectrometer withoutpretreatment. Smoothing, derivatives, vector normalization (VN) and principal component analysis (PCA) were employed to improve the performance of classification models. Then, the classification models were built by support vector machine (SVM). Experimental results show that both smoothing and VN can lead to a remarkable improvement while five-point smoothinghas the best performance increasing prediction accuracy (PA) from 91.11% to 96.11%. The overall results demonstrate that spectral preprocessing plays an important role in NIR spectroscopy and low-resolution NIR spectrometer together with appropriate chemometrics methods can be successfully applied in green tea origins identification.