[Objective] Representative purple soil and zonal yellow soil in Southwest China were selected to analyze their spectral information and to estimate soil moisture content in order to provide a method basis for rapid soil moisture monitoring in Southwest China.[Methods] Different soil moisture content levels were established in two soil types in the laboratory, and spectral reflectance was measured by using a ground surface spectrometer. The hyperspectral characteristics were compared and analyzed, and the characteristic bands were extracted by various mathematical transformations and correlation analysis. Hyperspectral estimation models of soil moisture were then constructed by stepwise multiple linear regression (SMLR) and BP neural network (BPNN).[Results] ① The spectral reflectance of both purple soil and yellow soil decreased as soil moisture content increased, and the spectral reflectance of purple soil was lower than that of yellow soil under the same soil moisture content. ② The effect of soil moisture content on the reflectance of infrared wavelengths (760-2 500 nm) was stronger than the reflectance of visible wavelengths (380-760 nm), and there were obvious water absorption valleys near 1 400, 1 900 and 2 200 nm. ③ There was a strong correlation between spectral reflectance and soil moisture content of purple soil and yellow soil after mathematical transformation. ④ The soil moisture prediction model based on BPNN was superior to SMLR.[Conclusion] The BPNN model was the best model for estimating soil moisture content of purple soil and yellow soil in Southwest China. The BPNN model can quickly and accurately obtain soil water status of purple soil and yellow soil.