Research On Near-infrared Modelling And Its Application In The Blend Threshing Of Tobacco Leaf

The traditional chemical detection operation is complex, time-consuming, high cost, large pollution, compared with traditional chemical detection operation, near-infrared spectroscopy analysis without sample pretreatment, no pollution, convenient, low cost and simple operation analysis is being widely applied in various fields. In our study,872tobacco samples were collected from9areas of Zunyi city to use for the modeling of chemical components, including total sugar, reducing sugar, establishment of nicotine, potassium and chlorine, nitrogen. The samples were spectral filtered and pretreatmented by Antaris type ft NIR spectrometer. Of six chemical components’model, the correlation coefficient and cross validation mean variance is:total sugar0.9823,1.19; reducing sugar0.9758,1.14; potassium0.9580,0.176; nicotine0.9766,0.252; nitrogen0.9623,0.103; chlorine0.9415,0.0539. Though external data cross test, the relative error and absolute error of the established models meet the experiment requirements, accuracy of the model is satisfactory.In this paper, using comparison analyses between near-infrared model and continuous flow analyzer, the detecting precision of different formula and blend threshing with different tobacco leaves was determined, including the upper and lower orange leaves; upper and lower variegated leaf; the upper orange leaves, variegated leaves and lower orange leaves, variegated leaves. The statistical results, the average relative deviations of nicotine and total sugar were4.51%and4.78%respectively, less than5%. There was no significant difference between near infrared model and continuous flow analyzer for the detection of mixed samples.