Research On The Classification Method Of Near-infrared Spectroscopy Based On The Origin Of Torreya Grandis

Torreya fruit nutritious, has a special smell phenol incense.In the dried fruit market, the price is relatively high Torreya, so a lot of unscrupulous businessmen sell home-grown or sold inferior Torreya grandis, but With the well-known "regional identity" brand bags packaging products.they sell seconds at best quality prices and disrupt the normal market order Torreya. The traditional method is time-consuming. In order to quickly identify with the "certification of origin" logo Torreya, this is the first use of near infrared spectroscopy combined with stoichiometry of different origin Torreya discriminant analysis research.In this study, the near-infrared diffuse reflectance spectroscopy acquisition Linan, Shaoxing,Shengzhou Torreya sample of each 100, and reject abnormal Torreya.The samples were divided into calibration set and prediction set by sampling Random(RS) in the 3:1 ratio, The calibration sample is84, and the forecast sample is 28.In the 4000~10000cm-1 band, the spectra of the first order derivative(FD) and multiple scattering(MSC) were selected, PCA-DA, KNN, PLS-DA, LS-SVM and SIMCA are used to establish the discriminant model, and comparative modeling effect of these five kinds of models.The results showed that: least squares support vector machine(LS-SVM) and principal component(PCA-LDA) modeling method has the best effect. in the LS-SVM model, The correct recognition rate of FD and MSC+FD was 100%, the correct recognition rate was 96.43%; in the PCA-LDA model, The correct recognition rate of the correct set is 98.81%, and the correct recognition rate of the test set is100%; in PLS- DA model, The correct recognition rate of the modeling set is not less than 90%, and the correct recognition rate is 100% for the MSC+FD and FD preprocessing methods; In the model of SIMCA and KNN,The average correct recognition rate of the modeling set and the test set is lower than70%.Experimental results show that: the use of near-infrared spectroscopy combined with LS-SVM pattern recognition or PCA-LDA method Torreya origin classification is feasible.