基于高光谱成像技术对甘草饮片中甘草苷和甘草酸含量快速无损检测*

作者：殷文俊1，彭 攀1，陈 星1，李 杰1，张 慧2

单位：1.株洲市中医伤科医院，湖南 株洲 412000； 2.浙江工业大学药学院，浙江 杭州 310014

引用：引用：殷文俊，彭攀，陈星，李杰，张慧.基于高光谱成像技术对甘草饮片中甘草苷和甘草酸含量快速无损检测[J].中医药导报,2026,32(5):61-68.

DOI：10.13862/j.cn43-1446/r.2026.05.010

PDF： 下载PDF

摘要：目的：利用高光谱成像技术结合预处理算法和特征波段筛选方法建立模型，对甘草饮片中甘草苷和甘草酸含量快速无损监测。方法：本研究利用高光谱成像技术对短波红外波段下的甘草饮片进行采集，通过高效液相色谱获取甘草饮片的甘草苷和甘草酸含量信息作为参考值，与样品光谱信息相对应。对原始光谱数据进行预处理以消除仪器的噪声干扰，依据甘草苷和甘草酸含量信息从特征波长和特征波段两个方面筛选特征波长，分别建立甘草苷和甘草酸含量的偏最小二乘回归（PLSR）、多元线性回归（MLR）和贝叶斯岭回归（BRR）定量校正模型，用预测集决定系数（R2p）和预测集误差均方根（RMSEP）作为模型性能的评价指标。结果：高光谱甘草苷和甘草酸定量模型最优预处理方法分别为Savitzky-Golay平滑和SNV；通过对特征波段方法的筛选，CARS方法能够筛选得到与甘草苷和甘草酸含量最为相关的波长；针对甘草苷含量建立的PLSR模型和甘草酸含量建立的BRR模型对含量预测具有最优精度，模型预测集决定系数（R2p）分别为0.958 7和0.897 3，预测集误差均方根（RMSEP）分别为0.001 8和0.003 8，表明所建立含量预测模型具有可靠的稳定性与准确性。结论：高光谱成像技术可实现甘草中甘草苷和甘草酸成分含量的快速有效检测，为中药甘草品质的快速无损检测提供了一种新的途径。

关键词：高光谱成像；甘草；甘草苷；甘草酸；无损检测

Abstract：

Objective: To establish the mathematical models for a rapid and nondestructive analytical method to monitor the content of liquiritin and glycyrrhizic acid in licorice. Methods: The contents of liquirtin and glycyrrhizic acid in licorice determined by High Performance Liquid Chromatography were set as the reference values, which associated with licorice spectra by using hyperspectral imaging technology in short-wave infrared ranges. In order to eliminate the possible instrument noise during the hyperspectral acquisition, the preprocessing methods of the spectra were carried out. Using the correlation between the content of liquirtin and glycyrrhizic acid and the reflectance data to select the effective band, quantitative calibration models were developed by partial least squares regression (PLSR), multiple linear regression (MLR) and bayesian ridge regression (BRR), and evaluated by the metrics of coefficients of determination of prediction (R2p) and the root mean square errors of prediction (RMSEP). Results: The results demonstrated that the optimal preprocessing methods for hyperspectral-based quantification of liquirtin and glycyrrhizic acid were Savitzky-Golay smoothing and standard normal variate transformation (SNV), respectively. Through feature band screening, the competitive adaptive reweighted sampling (CARS) algorithm effectively identified wavelengths exhibiting the strongest correlation with liquirtin and glycyrrhizic acid contents. Specifically, PLSR model developed for liquirtin content prediction and BRR model for glycyrrhizic acid content prediction achieved the highest predictive accuracy, with R2p of 0.958 7 and 0.897 3, and RMSEP of 0.001 8 and 0.003 8. These findings indicate that the established content prediction models exhibit robust stability and accuracy. Conclusion: Hyperspectral imaging technology can achieve rapid and effective detection of the content of liquirtin and glycyrrhizic acid in licorice, which provides a new approach to the rapid and nondestructive detection of the quality of licorice.

Key words：hyperspectral imaging; licorice; liquirtin; glycyrrhizic acid; nondestructive detection

发布时间：2026-05-23

点击量：387