Because of the similar appearance and properties of different quality grades of the product, super Dendrobium huoshanense could be easily adulterated with first-grade D. huoshanense and second-grade D. huoshanense products, thereby affecting its clinical application and causing market distortion. In this study, a combination of hand-held near infrared spectroscopy and chemometrics was used to classify different grades of D. huoshanense. The standard normal variate was employed to preprocess the original near infrared spectra, following which linear analysis models (principal component analysis (PCA), linear discriminant analysis (LDA), partial least squares discriminant analysis (PLSDA), and a non-linear support vector machine (SVM) model, were utilized to establish the identification models. The results showed that PCA analysis could not identify the three grades of D. huoshanense, and the LDA analysis could distinguish the second-grade from the other two grades. The PLSDA model resulted in prediction accuracies for the calibration cross-validation, and test sets of 91.83%, 83.58%, and 84.29%, respectively. Unfortunately, the super and first-grade D. huoshanense were not identified by the linear analysis model. Further analysis was performed with a non-linear model, where SVM was used to analyze all grades of D. huoshanense. The recognition rate of thel training set and validation set were 88% and 84%, respectively. All in all, the use of a hand-held near infrared spectrometer combined with chemometrics could identify the quality grade of D. huoshanense samples on-site in real-time, and provide a simple, fast, and reliable method for the quality control of the traditional Chinese medicine herb of D. huoshanense.

由于不同质量等级的产品外观性质相似，超级霍山石斛极易与一级霍山石斛和二级霍山石斛产品掺假，影响临床应用，造成市场扭曲。本研究采用手持近红外光谱和化学计量学相结合的方法对不同等级的霍山毛茛进行分类。. 使用标准正态变量对原始近红外光谱进行预处理，然后使用线性分析模型（主成分分析（PCA）、线性判别分析（LDA）、偏最小二乘判别分析（PLSDA）和非线性支持向量采用SVM模型建立识别模型，结果表明PCA分析不能识别出三个等级的火山毛茛，LDA分析可以区分二级和其他两个等级。PLSDA模型导致校准交叉验证和测试集的预测准确率分别为 91.83%、83.58% 和 84.29%。不幸的是，超级和一级D. hoshanense没有被线性分析模型识别。使用非线性模型进行进一步分析，其中 SVM 用于分析D. huoshanense的所有等级。训练集和验证集的识别率分别为88%和84%。总而言之，使用手持式近红外光谱仪结合化学计量学，可以在现场实时识别霍山毛茛样品的质量等级，为霍山毛茛的质量控制提供一种简单、快速、可靠的方法。 D. hoshanense的中药材。