Abstract: Loess soil samples (129) taken from the field in Qian county of Shaanxi Province in 2014, were chosen as objects. The gravimetric moisture content and spectra of soil samples were measured during the air drying process. The moisture absorption characteristic parameters extracted using the continuum-removal method, quantitatively predicted the moisture content in loess. Quantitative inversion models of soil moisture content were devised using linear regression, logarithm, power law, and exponential analysis based on analyzing the correlation between spectral absorption feature parameters and the soil moisture content. Results showed that the spectral absorption characteristic parameters of the most correlated coefficients with soil moisture were the maximum absorption depth (D) and total absorption area (A), absorption peak right area (R_A), and absorption peak left area (L_A). The correlation between the spectral absorption characteristic parameters and the soil moisture content in 1 900 nm was better than 1 400 nm. The best prediction models of soil moisture content were the linear models using the maximum absorption depth in 1 900nm (D_{1 900}) and the absorption peak right area in 1 900nm (R_{A1 900}) as independent variable and logarithm models using the total absorption area in 1 900 nm (A_{1 900}) and the total absorption area in 1 400nm (A_{1 400}) as the independent variable, respectively. The coefficient of determination (R²) of calibration and validation was bigger than 0.92 and 0.95, respectively, and the residual prediction deviation (RPD) was greater than 4, and the root mean squared error of prediction (RMSEp) was smaller than 1.5%. The study provides a reference for rapid determination the soil moisture in precision agriculture.