烟叶成熟度区分及烟碱含量检测模型研究

为探索一种有效的近红外光谱应用于烟叶成熟度区分以及烟碱含量检测模型建立的方法，对56组烟叶样品近红外光谱数据进行系统聚类，分类结果与原始结果基本一致，客观反映了成熟与未成熟烟叶的差异。识别率达到92.86%。采用多种不同的光谱预处理方法，并选择较优的多元散射校正处理原始光谱，再用偏最小二乘回归建立模型。结果表明：所建模型训练集r=0.9852，RMSE=0.0676；交叉验证r=0.9145，RMSE=0.1645。预测值能够较为均匀紧密地分布于拟合线的两侧，预测结果较好。

Study on Distinguishing the Maturity of Tobacco Leaves and Detection Model of Nicotine Content

An effective way that can distinguish the maturity of tobacco leaves and build a detection model of tobacco nicotine content using near infrared spectroscopy was explored. The NIR spectral date about 56 groups of tobacco leaves were analyzed by using clustering analysis, the classification results were almost the same as original, which reflected the difference between mature and immature tobacco leaves objectively. The rates of identification were up to 92.68%. A variety of spectroscopic processing methods were used to deal with the original spectrum. The method of multiplicative scatter correction (MSC) was better than others. Partial least squares regression (PLS) was used to established the regression model, which had the regression coefficient r=0.9852, RMSE=0.0676 in training set, r=0.9145, RMSE=0.1645 in cross validation set. The predicted values distributed closely in both sides of the fitting line, which showed a good result.