近红外光谱(Near-infrared spectroscopy，NIRS)技术作为一种快速无损检测方法，在许多领域广泛运用，尤其是在食品领域中的应用更加广泛。信阳毛尖茶是我国十大名茶之一，品质优异。本文通过采集三种不同等级信阳毛尖茶800~2500 nm处近红外光谱信息，对三个不同等级信阳毛尖茶的所有波长的响应进行单因素方差分析，选择了数据分析的波长范围；通过小波变换滤噪，对原始光谱进行预处理，采用正交偏最小二乘判别分析(OPLS-DA)和偏最小二乘法(PLS)对信阳毛尖茶品质进行判别。结果表明茶样品800 nm~1800 nm波长范围的近红外光谱数据可用于预测信阳毛尖茶品质；OPLS-DA分析表明三个不同等级信阳毛尖茶可以有效区分；所建立的PLS预测模型，理论值和预测值之间具有良好的相关性，相关系数为0.994，预测准确率为100%，交叉验证均方根误差(RMSECV)为0.090，表明模型预测准确、可靠。本研究建立的NIR光谱结合PLS分析方法可以用于快速无损检测河南信阳毛尖茶的等级品质。

Near-infrared spectroscopy technology was widely used in many fields as a kind of fast nondestructive testing method, especially, which was more extensively used in the field of food science. As one of top ten Chinese green teas, Maojian tea is also well known abroad with high quality. In this study, near-infrared spectral data from 800 to 2500 nm for three different grades of Maojian tea were collected and analyzed. One-way analysis of variance of near-infrared spectral data was acquired from the three different grades of Maojian tea, in order to choose the most suitable wavelength range of near-infrared spectral data for further analysis. The original spectral signals were pretreated by wavelet transform for denoising. Orthogonal partial least-squares discriminant analysis (OPLS-DA) and partial least squares (PLS) were employed to distinguish the quality grades of Maojian tea. Results showed that the near infrared spectral data obtained in the wavelength range from 800 nm to 1800 nm can be used to predict the quality grades of Maojian tea. The three different grades of Maojian tea can be distinguished effectively by OPLS-DA analysis. PLS forecast model was established, the theoretical values and predicted values showed good correlation, while the correlation coefficient is 0.994, prediction accuracy is 100%, Root mean square error of cross validation is 0.090, which showed the accuracy and reliability of the models. NIRS combined with PLS analysis can be used for fast nondestructive testing of the quality grades of Maojian tea.

河南省豫南茶树资源综合开发重点实验室开放基金资助项目(HNKLTOF2017007)；江西省教育厅科学技术研究项目(GJJ170291)资助