Establishment of Near-infrared Prediction Model for Seven Chemical Components of Wrapper Tobacco

Abstract

【Objective】It is complex, time-consuming and laborious for conventional chemical detection method to detect the content of total nitrogen, potassium, total sugars, reducing sugars, total alkali, chlorine and magnesium in wrapper tobacco, while near-infrared spectroscopy technology is simple to operate and rapid to detect. The aims is to establish a near infrared spectroscopy detection model to achieve rapid quantitative analysis of the content of the seven chemical components in wrapper tobacco.【Method】Taking Yunnan cigar wrapper tobacco as the material, conventional chemical method was used to detect the content of seven chemical components in wrapper tobacco, and then near infrared spectroscopy was used to analyze the content of seven chemical components and their spectruml data of wrapper tobacco by using near infrared spectroscopy combined with partial least squares method, and the model with the optimal predictive performance was determined by comparing the root mean square error and correlation coefficient of the model.【Result】The optimal prediction results of the seven chemical components were established by the preprocessing methods of original spectrum, first derivative, first derivative, original spectrum, original spectrum, first derivative + median filtering and first derivative + median filtering, and the optimal principal component numbers were 20, 7, 4, 24, 21, 9 and 7, respectively; The training set correlation coefficients (r) of the seven models were 0.9441, 0.8589, 0.7664, 0.9511, 0.9547, 0.9031 and 0.8620, respectively. The root mean square error of cross validation (RMSECV) were 0.1288, 0.2846, 0.0280, 0.0096, 0.1894, 0.2965 and 0.0795; The correlation coefficients (r) of the validation set were 0.8958, 0.7675, 0.7181, 0.7928, 0.7282, 0.8062 and 0.7980, and the root mean square error of prediction (RMSEP) were 0.1789, 0.3011, 0.0324, 0.0193, 0.3855, 0.3990 and 0.0999. Moreover, the models were externally verified, and the results indicated that the average RSD of the predicted values and chemical values of the seven chemical components were less than 32%.【Conclusion】It was feasible to use near infrared spectroscopy to rapidly and quantitatively analyze the contents of seven chemical components in wrapper tobacco. The models had good prediction effect on the contents of seven chemical components, and could provide a reference for the rapidly and quantitatively analysis of seven chemical components in wrapper tobacco.

Keywords

• wrapper tobacco
• chemical components
• near infrared spectrum
• preprocessing method
• partial least squares
• quantitative analysis