Gelatinization is one of the most important processing properties of millet, which has an important influence on the processing performance and product quality of millet. In this study, a rapid non-destructive detection method for the gelatinization characteristics of millet was proposed based on visible/near infrared spectroscopy technology which was used to detect the gelatinization properties of millet without crushing millet granules. Firstly, the VIS/NIR diffuse reflectance spectrum of millet in the range of 370~1 020 nm was collected, and crushed the millet into millet flour subsequently, then seven gelatinization characteristics such as the peak viscosity (PV), trough viscosity (TV), breakdown(BD), final viscosity (FV), setback (SB), pasting temperature (GT) and peak time (PT) were determined by RAV rapid viscosity analyzer. After that, the reflectance spectrum was preprocessed by Savitzkye-Golay smoothing (SG), multivariate scattering correction (MSC) and first derivative method (1-D). Finally, combined different preprocessed spectra and millet gelatinization characteristics, the calibration sets and validations set of the samples were determined by Sample set partitioning based on joint x distances (SPXY), and then based on the Successive projections algorithm(SPA) the optimal wavelengths were selected, the multivariate linear regression (MLR) prediction models of millet gelatinization characteristic were established using optimal wavelength reflection spectral signal, and the validation set samples were used to verify the prediction accuracy of the MLR model. MLR models prediction results: for the parameters of PV, TV and SB, after MSC pretreatment, the MLR model established by 9, 17 and 18 optimal wavelengths was the best, the predicted correlation coefficients (R_p) were 0.934 7, 0.825 5 and 0.874 6 respectively, and the standard error of predicted residual (SEP) were 174.039 7, 67.220 3 and 74.281 8 respectively; for BD value, the MLR model with 14 optimal wavelengths after S-G pretreatment was the best with the R_p was 0.924 4, and the SEP was 178.020 1; in addition, for the FV parameters, after 1-D pretreatment, the prediction R_p and SEP of MLR model based on 16 optimal wavelengths were 0.853 1 and 132.166 7 respectively. The results show that it is feasible to detect the gelatinization properties of millet without crushing millet by using VIS/NIR diffuse reflectance spectroscopy combined with SPXY and SPA algorithm. This study provides a new technical means for millet products enterprises to determine the gelatinization characteristics of millet raw materials in the early stage of production rapidly and evaluate the processing quality of millet products, which has a high practical application potential.