基于近红外光谱分析技术结合化学计量学方法的初烤烟叶霉变预测研究

为建立基于烟叶麦角甾醇含量结合近红外光谱分析技术的初烤烟叶霉变预警模型，以2015年和2016年云南5个地区2个等级（B2F和C3F）初烤烟叶为研究对象，调节烟叶含水率为18%，在28℃，RH 80%条件下以30天为实验周期，进行烟叶霉变实验。每3天取一次样，采集近红外光谱数据并检测样品麦角甾醇含量。建立第0d初烤烟叶样品近红外光谱主成分监测模型并提取Hotelling T2统计量，预测第3天至30天初烤烟叶样品近红外光谱数据的Hotelling T2统计量，对比分析肉眼观察和近红外类模型对烟叶霉变的预警效果。结果表明:1）烟叶霉变过程中，麦角甾醇含量逐渐增加后逐渐降低，当肉眼可见时，麦角甾醇含量较初始值增加4.66~23.38倍；2）基于上述监测模型，13个霉变烟叶样品中，提前预警天数为6天的样品2个，提前预警天数3天的样品7个，当天预警的样品4个，7个未发生霉变烟叶在30天的监测周期内均未出现预警，预测准确率100%。以上结果表明该方法能方便快速地实现对初烤烟叶霉变的预警，具有较好的实用价值。 

Prediction model for flue-cured tobacco leaf mildew based on NIR spectroscopy combined with chemometrics

A near infrared spectroscopy (NIR) prediction model was built for tobacco mildew early warning based on content of ergosterol. Twenty leaf tobacco samples of two grades (B2F and C3F) in 2015 and 2016, from five regions in Yunnan province, were collected as research material. Mildew tests were conducted under artificial conditions (28℃, 80% relative humidity, and 18% of leaf water content). Tobacco leaves were sampled every 3 days, and ergosterol content was determined. Based on principal component analysis (PCA) and Hotelling T2 parameter, a prediction model was established. Early warning of tobacco mildew was compared between model prediction and unaided eye observation. Results indicated that (1) when mildew became macrosopic, ergosterol content of tobacco leaf increased by 4.66-23.38 folds compared with initial value. (2) Among the 20 samples, nine confirmed mildew pre-warning by 3-6 days, and four warned at the same day when mildew was macrosopic, while no mildew was found in the rest seven samples. It was concluded that the model can be used to predict flue-cured tobacco leaf mildew.