| In order to standardize the market of oil-tea camellia seed oils and safeguard the rights of consumers, a rapid and accurate method for identification of pressed and extracted oil-tea camellia seed oil was established. A large number of pressed and extracted oil-tea camellia seed oil samples were scanned by Fourier transform infrared spectroscopy to extract the characteristic band data. Savitzky-Golay smoothing (SG), multivariate scatter correction (MSC), standard normal transformation (SNV), first derivative (FD) and second derivative (SD) methods were used to preprocess, then combined with partial least squares (PLS), support vector machine (SVM) and BP artificial neural network (BPANN) to establish identification model. The results showed that when BPANN and PLS were used to establish the identification models, the results of SG were the best, and the correlation coefficient of validation (RP), the root mean square error of validation (RMSEP) and the identification accuracy of the SG-PLS model and SG-BPANN model were 0.767 9 and 0.921 2, 0.322 6 and 0.205 9, 88.46% and 100% respectively. The SNV was the optimal preprocessing method for SVM modeling, and the RP, RMSEP and the identification accuracy of the SNV-SVM model were 0.761 4, 0.882 1 and 88.46% respectively. Therefore, infrared spectroscopy could be applied to the identification of pressed and extracted oil-tea camellia seed oils. |
