Soil water is an important component of the global ecosystem. Quantitative remote sensing estimation of soil water content in Karst Rocky Desertification Area can provide basic data and theoretical support for rocky desertification control and ecological restoration. It also provides guidance for agricultural activities in Rocky Desertification Areas. Based on Sentinel-1A and Landsat 8 image data, the backscatter coefficients of shrub land and sparse woodland were extracted by using water cloud model, and TVDI of dry land and forest land were calculated by simplified Ts/NDVI feature space. Combined with the measured data, the soil moisture content of different depths was modeled by fitting analysis, which was used to inverse the soil moisture content. The results show that the VH polarization quadratic curve model and the VH polarization cubic curve model are suitable for inversion of soil water content at depths of 0~5 cm and 5~10 cm in shrub lands, respectively. The R2 and RMSE of the two models were 0.87, 0.87 and 4.57%, 4.29% individually. The exponential regression model of VH polarization was applied to soil moisture inversion of sparse woodland in 0~5 cm depth and the linear regression model of VH polarization was suitable for 5~10 cm depth. The R2 and RMSE of the two models are 0.736, 0.72 and 9.77%, 11.28% respectively. The best soil moisture inversion models of dry land and forested land are the cubic curve model and the logistic regression model respectively. And the R2 and RMSE of 0~5 cm depth soil moisture inversion are 0.85, 0.69 and 2.88%, 4.02%, while in 5~10 cm depth the value of R2 and RMSE are 0.76, 0.23 and 3.5%, 6.37% individually.