Abstract: The appearance and quality of tea in seasons of spring, summer and autumn are quite different. Using vegetation index to monitor the free amino acid content of fresh tea leaves in different seasons is facing great challenges. The spectral transformation played an important role in highlighting the characteristic spectrum and eliminating the influence of background and noise. Whether the optimization of vegetation index（VI） by transformation was beneficial to the free amino acid content of tea leaves at multiple growth stages was concerned. The free amino acid content data and hyperspectral data of ten tea varieties（summer tea, autumn tea and spring tea） were analyzed in three consecutive seasons. Firstly, the original spectral data was transformed by spectral transformations(reciprocal(T_1/R), logarithm(T_{1 gR}), first-order differential T_R′, first-order differential T_（1/R）′ of reciprocal and first-order differential T_{（1 gR）′} of the logarithm). The correlation between the vegetation index of different transformation spectra and the combination of spectral transformation and the amino acid of fresh tea leaves in different seasons was further analyzed. Finally, the effects of different spectral transformations on the free amino acid model of fresh tea leaves in different seasons were compared. The results showed that the changing trend of free amino acid content in modeling set and validation set of fresh tea leaves was spring tea free amino acid content（modeling mean: 4.03%, validation mean: 3.98%）, autumn tea free amino acid content（modeling mean: 3.72%, validation mean: 3.56%） and summer tea free amino acid content（modeling mean: 2.91%, validation mean: 2.93%）. Except for T_lgR-TCARI, the correlation between other vegetation indices optimized by spectral transformation and free amino acids in fresh tea leaves was higher than that between classical vegetation indices and free amino acids in fresh tea leaves, with the absolute correlation coefficients of 0.10～0.30. The accuracy of the MLR model based on T_lgR-VI was obtained in the calibration sets and verification sets of different seasons, and it was suitable for the estimation of the amino acid content of tea fresh leaves during multiple growth periods. The multiple linear regression（MLR） model based on T_lgR-VI had high accuracy, with determination coefficient（R~2） of 0.38 and root mean squared error（RMSE） of 0.72% for calibration sets and R~2 of 0.38 and RMSE of 0.84%, respectively. The overall results indicated that spectral pretreatment had a positive effect on the monitoring of free amino acids in different growth seasons, which provided a technical reference for the estimation of tea quality.