Abstract: The conventional Ground Penetrating Radar (GPR) imaging algorithms are data-independent, which cause lots of interference like side lobes and clutter. However, the data-adaptive methods have remarkable capability of interference suppression. A data-adaptive GPR imaging algorithm based on the theory of Robust Capon Beamforming (RCB) is proposed. The refraction phenomenon of electromagnetic wave is fully considered, and the generation of covariance matrix is also deeply discussed. The sidelobes and clutter can be suppressed dramatically, as well as the imaging resolution can be slightly improved. The experimental results show the validity of the proposed algorithm.