基于高光谱成像的主成分分析和波谱角分类技术的天麻粉质量标志物及相关成分可视化快速检测

Visual rapid detection of quality markers and related components of Gastrodia elata powder based on principal component analysis and spectral angle mapper classification technology with hyperspectral imaging

中文关键词:  天麻粉  产地  加工  高光谱  质量标志物

英文关键词:Gastrodia elata powder  origin  processing  hyperspectral  quality markers

中文摘要:

      目的 利用高光谱成像(hyperspectral imaging, HSI)技术, 通过一种原位质量标志物评估方法, 与高效液相色谱含量测定相结合, 建立预测模型对不同来源的天麻粉进行可视化评价。方法 采用高效液相色谱法测定不同产地、不同晾干方式天麻粉中6种质量标志物含量并作为金标准; 苯酚-硫酸法测定天麻多糖含量、水分测定仪测定含水量, 两者作为辅助评价指标。HSI选用可见近红外(visible near-infrared, VNIR)和短波红外(short wave infrared, SWIR)双波段光谱进行分析比较。结果 以鲜品切块晾干方式制备的天麻粉的质量标志物对羟基苯甲醇、多糖和水分含量均增加, 同时天麻素和巴利森苷含量显著减少甚至消失。根据高光谱图像主成分分析(principal component analysis, PCA)预处理特征, 基于波谱角(spectral angle mapper, SAM)的监督分类算法对不同产地和加工方式制备的天麻粉进行了原位、无损分类, 其建立的线性回归模型实现了质量标志物含量的反演预测。结论 通过高效液相色谱含量测定的准确性和高光谱原位预测的无损性相结合的方式进行现场可视化评价和预测, 可以为不同产地、不同加工方式的天麻粉质量标志物快速检测提供技术参考。

英文摘要:

      Objective To establish a predictive model for visual evaluation of Gastrodia elata powder from different sources by an in-situ quality marker evaluation method utilizes hyperspectral imaging technology (HSI) combined with high performance liquid chromatography content determination. Methods High performance liquid chromatography was used to determine the content of 6 kinds of quality markers in Tianma powder from different regions and different drying methods, and used as the gold standard; the phenol sulfuric acid method was used to determine the content of polysaccharides in Gastrodia elata, and the moisture content was determined by a moisture meter, both of which were used as auxiliary evaluation indicators. HSI uses visible near infrared (VNIR) and short wave infrared (SWIR) dual band spectra for analysis and comparison. Results The quality markers of Gastrodia elata powder prepared by cutting fresh products and drying them increased the content of hydroxybenzyl alcohol, polysaccharides, and water, while the content of Gastrodia elata extract and Balisenside significantly decreased or even disappeared. Based on the preprocessed features of principal component analysis (PCA) in hyperspectral images, a supervised classification algorithm based on spectral angle mapper (SAM) was used to classify Gastrodia elata powder prepared from different regions and processing methods in situ and non-destructive manner. The linear regression model established by this algorithm achieved the inversion prediction of quality marker content. Conclusion By combining the accuracy of content determination by high performance liquid chromatography with the non-destructive in-situ prediction by hyperspectral analysis, on-site visualization evaluation and prediction can be carried out, providing technical reference for the rapid detection of quality markers of Gastrodia elata powder from different regions and processing methods.