Based on Landsat5 TM and ENVISAT-1 ASAR data, the effect of vegetation on the backscatter coefficient was removed by water-cloud model, and the surface soil moisture in the middle reaches of Heihe River was simulated using AIEM model. By establishing the relationship between vegetation water content(Mveg; the key parameter of water-cloud model) and multiple vegetation indexes, it was found that the accuracy of vegetation water content inversion by RVI was higher (R2=0.71). After removing the vegetation effect, the LUT table was established with AIEM model, and the soil moisture is estimated by the look-up table method based on the VV and VH polarization backscattering coefficients. Comparing the inversion results with field measured data, the study suggests that the VV inversion result is better (R2=0.74). Considering the influence of vegetation on radar backscatter coefficient, the proposed method is more suitable for inversion of soil moisture in vegetation-covered area.