A rapid detection method of using the near-infrared spectra（NIRS）to determine the concentration of syrup (trehalose, maltose and glucose ) in producing trehalose by synthase was established. The 65 NIRS samples of trehalose syrup were divided into a calibration set of 48 samples and a prediction set of 17 samples. Calculation results showed that the first derivative and savitzky-golay filter pretreatment is optimal, and the partial least squares (PLS) was more robust and predictive than the principal component regression (PCR) in building model by TQ analyst. Using PLS, 1D and savitzky-golay filter combination pretreatment model could not only reduce the background noise of the spectrum, but also improve the robustness of the model. The RMSEC, C_c and RMSEP, C_p of trehalose syrup showed in order that the trehalose model was 0.188, 0.995, 0.089, 0.989; the maltose model was 0.143, 0.997, 0.131, 0.969; the glucose model was 0.147, 0.997, 0.094, 0.999. The NIRS detection was fast and non-destructive, could be used to detect the composition of the trehalose syrup in producing it by synthase.