Tobacco Biochemical Component Spectrum Diagnosis Mechanism And Method

Spectral analysis technique, for its advantages of convenience, celerity, high precision and non-destroy measure, has beening applicated on many trades such as agricultural products, food, medicine, tobacco and so on. It is a hot-point of study and application in tobacco trade for high-speed inspecting the quality of fluecured tobacco leaf and its grade. The UV spectra,Fluorescence spectra and the near infrared spectra of different grades tobacco leaf are measured in this paper. The main biochemical component are also measured. The spectral differences between different fluecured tobacco leaves are compared. The correlation between their spectra and their contents of nicotine, total N and lipid were analyzed. The correlation between the grades of fluecured tobacco leaves and their spectra were also analyzed. Some UV spectal prediction models for the contents of nicotine, total N and lipid of fluecured tobacco leaves are established with different modeling methods by DPS software. The spectal prediction models for the contents of nicotine is better in 210～360nm than in other range. The contents of nicotine are very significantly correlative to the spectral parameters which the square of correlation coefficient R~2 = 0.8978 and F =30.7359. The spectal prediction models by the step regression method and PAC method for the contents of total N and lipid are better in 230～300nm, 270～300nm than in other range, respectively. The contents of total N and lipid are very significantly correlative to the spectra of fluecured tobacco which the square of correlation coefficient R~2 = 0.78723, F =2.9599 for total N and R~2 0.89174, F 28.8291 for lipid. The spectal prediction model of total N by PAC method is better than that by the step regression method. The relative error of the spectal prediction models of the lipid contents by the step regression method and PAC method are all less than 20%, which the error of 95% samples are less than 10%. The relative error of the spectal prediction models of the nicotine contents by the step regression method are commonly less than 10%. There are veracity for inspect models of the contents of nicotine, total N and lipid by UV spectrum. The results are basic consistent to that by biochemical method. This provides basis for analyzing the quality of fluecured tobacco leaf by UV spectrum, and that classifying fluecured tobacco leaf by spectral method.