目的 探究近红外光谱技术应用于川贝母Fritillariae Cirrhosae Bulbus快速辨识的可行性。方法 收集80个川贝母待测样品（含炉贝、松贝、青贝及川贝母伪品等），为获取上述样品的真伪和规格信息，首先进行传统人工鉴别（M₁法）与《中国药典》法鉴别（M₂法），并以M₁法和M₂法相结合的辨识结果作为标杆信息（Y）；近红外光谱仪采集待测样品粉末的光谱信息（X），结合主成分分析-判别分析（principal component analysis-discriminant analysis，PCA-DA）、偏最小二乘-判别分析（partial least squares discriminant analysis，PLS-DA）、最小二乘支持向量机（least squares support vector machines，LS-SVM）3种化学计量学方法建立并优化真伪及商品规格辨识模型[Y＝M(X)]（M₃法）。结果 80个样品均参与真伪辨识，经留一法交互验证，结果发现3种真伪辨识模型中均无未分类样品，准确率依次为98.75%、98.75%、97.50%；80个样品中有5个因专家对规格的判定意见难以达成一致或样品本身原因等不参与建模，最终以75个样品参与规格辨识，模型准确率分别为100.00%（无未分类样品）、100.00%（有4个未分类样品）、100.00%（无未分类样品）。结论 真伪辨识均以PCA-DA、PLS-DA为最终辨识模型，商品规格分类以PCA-DA、LS-SVM为最优辨识模型。表明近红外光谱技术在川贝母真伪和规格质量快速辨识方面具有良好应用前景。

Objective To explore the feasibility of the application of near infrared spectroscopy in the rapid identification of Fritillariae Cirrhosae Bulbus.Methods A total of 80 samples of Fritillariae Cirrhosae Bulbus were collected. In order to obtain the authenticity and specification information of the samples, traditional manual identification (M₁ method) and pharmacopoeia identification (M₂ method) were carried out, and the identification results combined with M₁ method and M₂ method were used as benchmark information (Y). Near Infrared Spectroscopy to collect spectral information of the sample powder (X), combined with principal component analysis, discriminant analysis (PCA-DA), partial least squares discriminant analysis (PLS-DA), the least squares support vector machines (LS-SVM) three chemometrics methods to establish and optimize the authenticity and commodity specification identification model [Y=M (X)] (M₃) Results All the 80 samples participated in the identification of authenticity and falsification. The results showed that there were no unclassified samples in the three identification models, and the accuracy rates were 98.75%, 98.75% and 83.75%, respectively. Among the 80 samples, five samples did not participate in the modeling due to the difficulty in reaching an agreement on the specification by experts or due to the sample itself. Finally, 75 samples were selected to participate in the specification identification, and the model accuracy rates were 100.00% (no unclassified samples), 100.00% (four unclassified samples) and 100.00% (no unclassified sample), respectively. Conclusion PCA-DA and PLS-DA were used as the final identification models for authenticity identification, and PCA-DA and LS-SVM were used as the optimal identification model for commodity specification classification. Near infrared spectroscopy has a good application prospect in the rapid identification of the authenticity and specification of Fritillariae Cirrhosae Bulbus.

R286.2

国家自然科学基金面上项目（81773892）；河南省卫生健康中青年学科带头人专项（HNSWJW-2020014）；河南省中医药拔尖人才培养项目（重点项目）（2019ZYBJ07）；河南省高层次人才特殊支持“中原千人计划”—“中原青年拔尖人才”项目（ZYQR201912158）；中管局基地专项（2018JDZX087）