The harvest time of traditional Chinese medicine (TCM) is a very essential part for the production and quality of TCM which is the prerequisite for safe and effective clinical use of TCM. It is of great importance to carry out the research of timely harvest time of TCM. With Fourier transform infrared spectroscopy (FTIR) to study harvest time of Seventy-two Gentiana Rigescens samples. First derivative, second derivative, standard normal variate, multiplicative scatter correction and Savitaky-Golay(15,3) smoothing of all original spectra were pretreated with TQ8.0 software. Samples were divided into calibration set and prediction set at the ratio of 3∶1. Principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA) model were established. The result indicated that after removing noise spectrum, the spectra range was from 1 800 to 600 cm-1, the method SNV combined with SD and SG present the best result of spectra pretreatment. The contribution rates of first three principal components were 92.47% with PCA. Small differences were found for the samples harvested in May, September and October. Same spectrum range was chosen and PLS-DA was applied to establish the model. The R2 and RMSEE were 0.967 8, 0.086 0, respectively, and the prediction accuracy is 100%. The methods of PCA and PLS-DA have good ability to classify and identify different harvest time of Gentiana Rigescens. It provided a basis for the identification of different harvest time of TCM.