Study on Moisture Content Prediction of Tobacco Leaf Based on Near Infrared Spectroscopy

In order to realize rapid nondestructive testing of moisture content of fresh tobacco leaves, near infrared spectroscopy was used to collect fresh tobacco leaves.The effects of four pretreatment methods, Multiple scatter correction (MSC), Standard normal variate (SNV), Normalize and Autoscales, on the accuracy of the model were compared.It was found that the model established after pretreatment with MSC had the best effect, Model R
²
was 0.8145 and RMSE was 0.1185.Two feature band screening methods, continuous projection method (SPA) and competitive adaptive weighted weight algorithm (CARS), were selected for wavelength selection based on the spectral data pretreated by MSC.The results showd that the established MSC-CARS-SVM model has the best effect, with the R
²
was 0.8381 and the root mean square error was 0.1158.It was proved that the near infrared spectroscopy technique could be used to detect the moisture content of fresh tobacco leaf rapidly and nondestructively.