The methodology of RFM modeling for spaceborne SAR and its application in SAR image geocoding are investigated in this paper. The issues associated with model construction

parameter solution and ref inement of RFM model are extensively studied. When the rigorous sensor model is known

a virtual control point grid can be established in a terrain-independent way for RFM modeling. A new approach is implemented to obtain unbiased solutions of RFM model parameters at a low computation time cost. The resultant RFM model can be used as a reliable substitute for the Range-Doppler （RD） model for geometric processing of SAR images as the f itting accuracy with reference to the RD model is usually better than 0.01 pixels. However

in general there will be geometric biases in the geocoded SAR images because systematic absolute geolocation errors usually exist in the RD and associated RFM models. To solve this problem

control points are used to derive an additional mathematical transformation to remove the absolute geolocation errors in the RFM model. Experimental results with ENVISAT ASAR data show the effectiveness of the proposed method for RFM model construction.