中文摘要：

试验旨在研究应用近红外光谱技术快速测定烟草中烟碱、总氮、总糖含量的可行性.使用偏最小二乘法（PLS）为建模方法,选择4 030～9 000 cm-谱段,采用一阶导数和SG平滑滤波法进行光谱预处理,建立了烟草中烟碱、总氮、总糖的近红外预测模型.3种组分最佳PLS预测模型的相关系数r分别为0.993 2,0.963 5和0.9810,交叉检验的均方差（RMSECV）分别为0.076 7,0.077 2和0.984 0.考察了模型的精密度,并将近红外光谱技术与参考方法（YC/T161）相比较,用实际样品进行了验证.结果表明：所建模型精密度良好,两种方法测得值不存在显著性差异（P＜0.05）,近红外光谱技术可以准确地定量分析烟草中烟碱、总氮、总糖.

英文摘要：

Fourier transform near-infrared （FT-NIR） spectroscopy was used to determine nicotine,total nitrogen and total sugar in tobacco.Partial Least Squares （PLS） for the range of 4 030 ～9 000 cm-1,first derivative and Savitzky-Goly filter were used to forecast the contents of nicotine,total nitrogen and total sugar in tobacco.The correlation coefficients for nicotine,total nitrogen and total sugar were 0.993 2,0.963 5,0.981 0,respectively.The root mean square error of cross validation （RMSECV） were 0.076 7,0.077 2,0.984 0,respectively.The accuracy of the model was satisfactory.The results showed that the two methods have no significant difference （P 〈 0.05）.It could be used to determine the contents of nicotine,total nitrogen and total sugar in tobacco rapidly.