论文部分内容阅读
采用近红外光谱分析技术,对不同硫磺熏蒸的葛根进行鉴别研究。将葛根饮片及其硫熏饮片进行横截面与纵截面原始光谱的采集,并对原始光谱进行主成分分析(PCA),采用随机法划分样本集,进而比较不同预处理方法,通过对样品不同截面光谱建立判别模型,计算每个模型外部验证的识别率。采用正交性检验和非参数检验对数据进行正态性检验。结果表明不同截面由于结构不同所以吸光度不同,并且对原始光谱进行主成分分析,发现当主成分数为3时累计贡献率达到99.2%,MSC+1D+Savitzky-Golay预处理方法所建模型最好,判别模型中横截面的50个模型,其识别率为(94.4±0.66)%,纵截面的识别率为(95.6±0.75)%,总截面为(95.3±0.65)%。3组识别率的差异并不大。对各组模型的预测性能进行差异性分析,认为横截面、纵截面及总截面建模结果没有明显差异,说明近红外光谱分析技术可以应用于中药材硫熏与否的快速鉴别。
Near infrared spectroscopy was used to identify the Pueraria lobata root fumigated with different sulfur. The Radix Puerariae slices and its sulfur-smoked slices were collected for the original spectrum of the cross-section and the longitudinal section, and the principal component analysis (PCA) of the original spectra was performed. The random sample was used to divide the sample sets, and then different pretreatment methods were compared. Spectra Establish a discriminant model to calculate the recognition rate for each model’s external validation. The data were tested for normality using orthogonality tests and nonparametric tests. The results showed that the absorbance was different due to the different structures of different cross-sections, and the principal component analysis of the original spectra showed that the cumulative contribution rate reached 99.2% when the principal component number was 3, and the MSC + 1D + Savitzky-Golay pretreatment model was the best, 50 models of the cross section of the model were identified, the recognition rate was (94.4 ± 0.66)%, the recognition rate of longitudinal section was (95.6 ± 0.75)%, and the total cross section was (95.3 ± 0.65)%. The difference in recognition rate among the three groups is not significant. The differences of the prediction performance of each group of models were analyzed. It is concluded that there is no significant difference in the cross-section, longitudinal section and total section modeling results, which indicates that NIRS can be applied to the rapid identification of sulfur smoked or not.