| Lycium barbarum has been used as a traditional Chinese medicine and health food. Moreover, its functional characteristics were widely accepted by consumers. Due to ecological environment, climate and other factors, the quality of L. barbarum are quite different. The unique geographical and climatic conditions in Ningxia created Ningxia L. barbarum excellent quality, which has been included in the "Chinese Pharmacopoeia". L. barbarum grown in Ningxia Zhongning has the best quality in all regions of Ningxia. With the rapid development of L. barbarum industry, the phenomena of malicious fraud, adulteration and trademark abuse of origins become more serious, which seriously disrupt the normal market order. Traditional morphological characteristics and chemical identification of the origin and quality are easily influenced by internal and external factors. In addition, the operation is complicated and the cost is relatively high. Thus, it is of great importance to develop an efficient, rapid and accurate ananlysis method. Near-infrared(NIR) spectroscopy can be an excellent candidate for geographical origin discrimination and rapid detection of quality with advantages of no damage detection, simple operation, and no reagents. Therefore, near-infrared technology combined with chemometrics methods were employed for discrimination analysis to identify L. barbarum grown in different origins; meanwhile, they were employed for rapid determination of L. barbarum polysaccharide(LBP), total sugar and crude fat content to achieve the rapid identification of L. barbarum quality levels. The main research results were as follows:(1) In order to identify L. barbarum origins of Ningxia Zhongning, Ningxia non-Zhongning and non-Ningxia, the NIR spectra of 208 samples were collected and processed by chemometric methods. The results showed that the best spectral reproducibility was obtained when the pressure height is 10mm; samples can be clustered into three categories by principal component analysis with Savitzky-Golay Derivatives combined with standard normal transformation(SNV) in wavelengths range of 7400-4000 cm-1. Soft independent modeling of class analogy and random forests algorithm were used for geographical origin discrimination; the correct rates of blind validation were 90% and 88%, respectively, which indicated that NIR technology can be applied in L. barbarum origin discrimination.(2) L. barbarum were dried to constant weight. The dried samples were crushed into powders mechanically. Water, LBP, total sugar and crude fat content were determined. The results showed that the average of water content was 7.69%±0.48%, and the average of water content in Ningxia, Inner Mongolia and Hebei is higher in Gansu, Qinghai and Xinjiang. The LBP content in Ningxia Zhongning was close to that in Ningxia non-Zhongning; the average of LBP content of Ningxia L. barbarum is higher than the other regions. The average of total sugar content of all samples were more than 50 g in a 100 g dried medlar, of which Qinghai L. barbarum has the highest content, namely 59.53 g. Hebei and Ningxia have the higher average content, while Inner Mongolia and Qinghai have the lower average content.(3) Partial least squares(PLS) models of LBP and total sugar were set up based on the relationship between the diffuse reflection spectral information and chemical measurements to determine LBP and total sugar rapidly. The modeling performance with different spectral preprocessing, modeling bands and the principal component numbers(PCs) were compared. LBP modeling results showed that the model was built with pretreatment spectra of SNV combined with Savitzky-Golay Derivatives and effective wavelengths(EWs) of 12000-7463 cm-1, 7002-5568 cm-1, 5087-4003 cm-1 selected by correlation coefficient. The determination coefficients in calibration(RC2) and validation(RV2) sets were higher than 0.93 when PCs was 5. Total sugar modeling results showed that the model was built with pretreatment spectra of Norris Gap Derivatives and EWs of 8700-4000 cm-1 selected by correlation coefficient. The RC2 and RV2 were higher than 0.91 when PCs was 8. For PLS models of LBP and total sugar, both of correlation coefficient selection and X-loading weights can improve the model, performance and the former was superior to the latter.(4) NIR spectra were obtained in diffuse reflection pattern. PLS model of crude fat was set up based on the relationship between the spectral information and chemical measurements. Crude fat modeling results showed that the model was built with pretreatment spectra of Savitzky-Golay Smothing and full wavelengths of 12000-4000 cm-1. The RC2 and RV2 were higher than 0.94 when PCs was 8. For PLS model of crude fat, both of correlation coefficient selection and X-loading weights have not obvious improvement of modelling. |
PDF Full Text Request