Rapid Discrimination Varieties And Geographical Origins Of Rice By Using Raman Spectroscopy And Chemometrics

As a kind of traditional Chinese diet, rice is the staple food for more than 60% of the population in China. The qualities of rice are affected not only by the genetic characteristics but also by geographical environment and climate environment. Therefore, the qualities are obvious different among the rice cultivated in different regions or different varieties. Recently, the counterfeit or substandard rice was occurred in market for the not enough management system. Meanwhile, it costs a relative longer time for distinguishing the rice by the traditional method. In this research, the rapid qualitative identification analysis model of rice was established by combining the confocal Raman spectroscopy with chemometric methods, which could provide a simple and effective method for the classification of rice, its origin traceability and variety discrimination.Firstly, the optimal method for collecting the rice’s information was established by application of Raman spectrometer. The five influencing factor, including excitation wavelength, exposure time, scan times, magnification and the sacn range, were optimized as following: the excitation wavelength 632.8 nm, the magnification 50 times, the exposure time 30 s, the scan times 2 times, and the scan range 200-1600 cm-1. The optimal method was proved to have good reproducibility by testing the single rice or the rice in the same batch.Secondly, the effects of different data preprocessing methods on the rice’s Raman spectrograms were investigated in this study. The different preprocessing methods included spectral average, smoothing, derivative, baseline, normalization, standardized normal variate, multiplicative scatter correction and the mean center.Thirdly, the varieties, classifications and origins of rice were identified by chemometrics method on the basis of their Raman spectra. Principal component analysis(PCA) was carried out to make a preliminary classification of rice samples and the results showed that rice samples could be classified rightly by PCA. Through the comparison of the different pretreatment methods, the best model was determined. As for the classification of Japonica rice, indica rice and glutinous rice, the right rate of hierarchical clustering analysis(HCA) was 94.29% after the baseline correction and 9 point smoothing processing; the best partial least square discriminant analysis(PLSDA) model was established which R2 was 0.97 after baseline processing and the right rates of calibration set and prediction set were both 100%; the best soft independent modeling of class analogy(SIMCA) model was established with baseline processed and the recognition rate and rejection rate of the prediction set were 98.74% and 92.44% at the significant level of α = 5%. As for the classification of Japonica rice and indica rice growing areas, the right rates of HCA were 80% and 100% after the baseline processed; the best PLSDA models were established which R2 were 0.99 and 0.98 respectively after first derivative processed and the right rates of calibration set were both 100% and the right rates of prediction set were 95.92% and 97.40%; the best SIMCA model was established with first derivative processed and the recognition rates of the prediction set were 100% and 99.68% as well as the rejection rate were 87.97% and 99.03% respectively at the significant level of α = 5%. As for the discrimination of rice varieties, the right rate of HCA was 90% after the baseline correction and 9 point smoothing processing; the best PLSDA model was established which R2 was 0.99 after baseline with SNV processing, and the right rates of calibration set and prediction set were 100% and 92.73% respectively; the best SIMCA model was established with baseline processed and the recognition rate and rejection rate of the prediction set were 100% and 94.92% respectively at the significant level of α = 5%.Finally, rapid qualitative identification analysis model was verified by testing the rice ‘Daohuaxiang’, which had the higher acceptance by consumers. The Raman spectra informations of Daohuaxiang rice were collected. The SIMCA model and PLSDA model were established by the above process. The results showed that the right rates of calibration set were 100% for the two models. And the rates of identification and rejection of SIMCA model were 80% and 100%, respectively, for the unknown samples. The rate of identification of PLSDA model was 100%. Therefore, the results indicated the established models could be used for identifying the rice ‘Daohuaxiang’.