Based on measured vegetation canopy hyperspectral data and HSI hyperspectral image data in a typical region, sensitive bands to soil moisture in different soil depths were selected by correlation analyses to establish the optimal inversion model of soil moisture by HSI image data, and the inversion model was calibrated by HSI inversion model based on the measured soil moisture. The results showed that: the sensitive bands to soil moisture were in 450 – 650 nm and in 850 – 920 nm; The two soil moisture inversion models showed that: the estimation effect for soil moisture in 0 – 10 cm depth was the best, and the accuracy of the inversion model based on HSI image data was higher than the inversion model based on the measured hyperspectral data, and the coefficients of determination (R²) were 0.659 and 0.557, respectively; The accuracy of soil moisture inversion model was improved better after calibration, the coefficient of determination (R²) raised to 0.719 from 0.557, and root mean square error (RMSE) was of 0.043 5, which indicated that the improvement of monitoring accuracy on soil moisture at the regional scale and proved this technique is feasible to monitor soil moisture, and provided helps to further improve soil moisture monitoring by remote sensing at regional scale.