Active microwave remote sensing and passive optical remote sensing have their own advantages and disadvantages in inversion of soil moisture. In order to combine the advantages of both of them to make up for shortcomings, a semi-empirical model based on Radarsat 2 data and Landsat 8 data has been presented for vegetation-covered soil moisture inversion. The model is based on water-cloud model with the vegetation water content estimated by the optical remote sensing as the key input parameter. Thus the influence of vegetation on the backscattering coefficient would be reduced. Combination of Radarsat 2 and Landsat 8 data were used to estimate the vegetation-covered soil moisture with the new coupling model in the studied area located in Eerguna City of Inner Mongolia, west of Greater Khingan. Then it was verified with the ground survey data. The research showed that the precision is high in the retrieval of vegetation water content (R2=0.89) using Landsat 8 data. The inversion accuracy of coupling model is higher than former algorithms. The R2 of HH polarization is raised from 0.27 to 0.65. These results showed that the proposed coupling model has a better inversion accuracy, and can be used in the inversion of vegetation-covered soil moisture.