Tea polyphenol is the main component in tea. The content of tea polyphenols accounts for about 30%, which determines the taste and color of tea soup. Therefore, how to use the NIRS to quickly detect the content of tea polyphenols has a very high practical value in the rapid identification of tea quality. On the basis of the combination between NIRS and chemometrics methods, this paper analyzes the components of different teas. The tests indicate that, there is an approximately linear relationship between tea polyphenol content and the absorbance in the near-infrared waveband (1800~2500nm). Based on this result, a fitting curve was established, and the correlation coefficient and the correction root mean square error of different fitting curves were obtained. The tea polyphenol content prediction model was established using near-infrared light method and partial least squares method at 1872nm. The correlation coefficient reached 0.9378, the RMSE is 0.008015.