Nondestructive Detection Of Volatile Oil Content In Zanthoxylum Bungeagum Maxim By Near Infrared Spectroscopy

The volatile oil is the main aroma ingredient in Zanthoxylum bungeagum Maxim, and the volatile oil content is one of primary physicochemical properties to classify Zanthoxylum bungeagum Maxim. The traditional chemical methods to measure the volatile oil content encounter with some problems such as long time, low efficiency, destroying sample, so it is difficult to achieve rapid detection. Near infrared spectroscopy (NIR) is new, fast and nondestructive method and applied to agriculture and food widely. To detect volatile oil content nondestructively, NIR is applied in the quality detection of Zanthoxylum bungeagum Maxim.In this research, 138 samples were collected from many provinces in China such as Shan Xi, Si Chuan, Gan Shu, Chong Qing, Yun Nan. The NIR spectra were collected with 9 different kinds of combination of scanning parameters from 3600～12800 cm^-1, and the volatile oil content was measured according to the distillation stipulated by the Forestry Standard of PRC-Quality Classify of Prickly Ash(LY/T 1652-2005). Calibration models of volatile oil content were established based on NIR and partial least squares (PLS). Result showed the model at resolution of 16cm^-1 and sample scans of 128 was better than the others, and the determination coefficient, RMSEP and RPD of test set validation were 0.973,0.272 and 6.28, respectively by test set validation.Derivative, smoothing, multiplicative scatter correction (MSC), and standard normal variate (SNV), Direct orthogonal signal correction (DOSC) were compared. By 1st derivative + 9 points average smoothing, the model performance was improved, and RPD was 8.61. Result showed proper data preprocessing methods can eliminate the effect of spectra shift and noise. DOSC can not only improve the model, but also simplify it.Genetic algorithm (GA) was applied to spectral region selection. The wavelengths used to build model were reduced and the model was improved by Genetic algorithm. Genetic algorithm was applied to support vector machine (SVM) parameters selection. Firstly, SVM parameters were searched at large range, and then searched at small range by GA. Result showed that the method can remedy shortcoming of exhaustive method and improve performance of model.At last, the models based on PLS, BP neural network, and SVM were compared. The PLS was the best, and the linear method was better than nonlinear it. NIR can be used as a method to detect volatile oil content in intact Zanthoxylum bungeagum Maxim rapidly and nondestructively.