摘要
为实现对烤烟香型风格的快速判别,应用可见-近红外光谱技术对浓香型、中间香型、清香型3种烤烟样品进行光谱采集,获得烟叶内在化学成分的特征光谱信息。选取全波段特征光谱经均值归一化、Savitzky-Golay卷积平滑、多元散射校正和标准正态变换等优化预处理,采用主成分分析(PCA)对光谱数据进行聚类分析,并结合偏最小二乘判别分析(PLSDA)建立了3种香型风格烤烟定性判别模型。结果表明,建立的PLS-DA模型的校正集和验证集的正确识别率均达到100%,模型性能良好,说明利用可见-近红外光谱技术对不同香型风格烤烟具有很好的分类和鉴别效果,为实现烤烟不同香型风格的快速、准确判别提供了一种新方法。
In order to achieve rapid discrimination of different Flue-cured tobacco aroma types, visible and near-infrared spectroscopy was used to obtain the characteristic spectral information of intrinsic chemical composition of three types of flue-cured tobacco including the full, middle and clean aroma types. The full-wave band of original characteristic spectra were selected for spectral pretreatment by normalized treatment, Savitzky-Golay, multiplicative scatter correction and standard normal variate. The principal component analysis （PCA） was applied to perform the clustering analysis, and a qualitative discrimination model of the three flue- cured tobacco arama styles combined with partial least squares-discriminant analysis （PLS-DA） was built. The results showed that the calibration and validation sets of PLS-DA model were identified as having good performance with an accuracy of 100%. The results suggested that this method using visible and near-infrared spectroscopy is good for classification and identification of tobacco with different aroma types. This study provides a new method for rapid and accurate identification of different flue-cured tobacco aroma types.
出处
《中国烟草科学》
CSCD
北大核心
2015年第6期88-93,共6页
Chinese Tobacco Science
基金
中国烟草总公司项目“浓香型特色优质烟叶风格定位研究”(Ts-01-2011001)
关键词
可见-近红外光谱
烤烟
香型风格
主成分分析
判别分析
visible and near-infrared spectroscopy
flue-cured tobacco
aroma type
principal component analysis
discriminant analysis
