论文部分内容阅读
对杀青烟叶蛋白质和水溶性总糖含量的光谱检测,发现近红外光谱(1100—2500nm)的检测模型优于可见-近红外光谱(350—2526nm),烟粉检测模型优于片状烟叶的检测模型。通过对烟叶全部光谱数据不同的预处理来探究其蛋白质和水溶性总糖的近红外光谱的检测模型,并利用近红外有效波长对施木克值的含量进行预测。利用偏最小二乘法(PLS)通过训练集的交叉验证建立回归模型,结果表明:(1)对原始光谱进行二阶导数变换后,得到蛋白质含量预测模型的预测集r=0.9768、RMSE=0.6843;(2)对原始光谱每隔51个点进行移动平滑处理及主成分数为8时,水溶性总糖含量预测模型的预测集r=0.9495、RMSE=0.9049;(3)基于82个波长对施木克值的预测模型的预测集r=0.9356、RMSE=0.1060。
The results showed that the detection model of near-infrared spectroscopy (1100-2500nm) was better than that of visible-near-infrared spectroscopy (350-2526nm), and the detection model of tobacco smoke was superior to that of flake tobacco model. Through the pretreatment of the whole spectral data of tobacco leaves, the detection model of near-infrared spectra of protein and water-soluble total sugar was explored, and the content of Shigel’s grams was predicted by using near-infrared effective wavelength. The regression model was established by partial least square method (PLS) through cross validation of training set. The results showed that: (1) After the second derivative transformation of the original spectra, the prediction set of the protein content prediction model was r = 0.9768 and RMSE was 0.6843; (2) The predicted set of water-soluble total sugar content predictive models r = 0.9495, RMSE = 0.9049, when the original spectrum was moved and smoothed every 51 points and the number of principal components was 8; (3) The predicted set of mc values for the prediction model r = 0.9356, RMSE = 0.1060.