The reduction of background signal, or clutter, from Ground Penetrating Radar (GPR) measurements is an area of active research. The weak reflection signal obtained from subsurface targets is usually blurred by such strong clutter, which mainly comes from flat or rough ground surfaces, underground inhomogeneities, and coupling between the transmitting and receiving antennae. In this paper, the improved Kalman filter techniques have been studied and applied to reduce the background interference signals and detect the buried targets in GPR dataset. The effectiveness and validities of the proposed improvement methods in this paper for processing GPR detection data are studied. The processed results prove that the proposed methods are effective and adaptive for reducing clutter and detecting subsurface targets.