目的 建立重楼及其混伪品的近红外光谱（near-infrared spectroscopy，NIRS）快速鉴别方法。方法 采用NIRS仪获取样品光谱数据，结合主成分分析（principal component analysis，PCA）、正交偏最小二乘判别分析（orthogonal partial least squares discriminant analysis，OPLS-DA）、线性判别分析（linear discriminant analysis，LDA）、人工神经网络机器学习算法（artificial neural network，ANN）和二维红外光谱（2D-IR），探讨NIRS技术在重楼及其混伪品快速鉴别的可行性。结果 重楼及其混伪品的NIRS光谱吸收峰的峰形整体相似，但在峰数和峰强方面均存在一定差异。PCA和OPLS-DA判别模型可将重楼及其混伪品区分，但模型预测能力差（Q²＜0.5）；LDA模型可将滇重楼与其他物种进行区分，但无法区分七叶一枝花的部分样本；ANN模型识别准确率达100.0%，不同物种的2D-IR图谱在5 897～5 600 cm⁻¹和4 497～4 200 cm⁻¹差异显著。结论 重楼及其混伪品的化学信息差异明显，不可混用。NIRS结合ANN模型或2D-IR方法可用于重楼及其混伪品的快速鉴别。

Objective To establish a rapid identification method for Paridis Rhizoma using near-infrared spectroscopy (NIRS). Methods The spectral data of the sample were obtained by NIRS instrument. Subsequently, the feasibility of NIRS in identifying Paridis Rhizoma and its adulterants was explored by using principal component analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA), linear discriminant analysis (LDA), artificial neural network (ANN), and two-dimensional infrared spectroscopy (2D-IR). Results The original spectral images of Paridis Rhizoma and its adulterants exhibited similar shapes overall, but differences were observed in the number and intensity of peaks. Although the PCA and OPLS-DA discrimination model successfully separated Paridis Rhizoma from its adulterants, it displayed poor predictive ability (Q²< 0.5); The LDA model could differentiate Paris polyphylla var. yunnanensis from other species, but face challenges in distinguishing certain samples of P. polyphylla var. chinensis; In contrast, ANN model demonstrated a clear advantage in identifying Paridis Rhizoma and its adulterants with a recognition accuracy of 100.0%. Additionally, the 2D-IR spectra of different species were significant differences between 5 897—5 600 cm⁻¹ and 4 497—4 200 cm⁻¹ regions. Conclusion There are notable chemical distinctions between Paridis Rhizoma and its adulterants that necessitate differentiation for proper use. NIRS, combined with the ANN model or the 2D-IR method, can be effectively employed for the rapid identification of Paridis Rhizoma and its adulterants.

R286.2

云南省院士专家工作站（202205AF150026）；国家自然科学基金项目资助（31860080）