The full polarization RadarSat-2 data used in land cover classification requires a lot of research and demonstration. Taking NLCD data as reference, this paper uses the SVM to classify 1 scene of RadarSat-2 full polarization data in San Francisco and then evaluates the accuracy of land cover classification results from area of consistency and spatial similarity. Experiments for SAR data classification get the highest overall classification accuracy of 76.91% and kappa coefficient of 0.65, which show that the fully polarization RadarSat-2 data can achieve good classification quality when it is used for land cover classification.