Space-time adaptive processing is an effective method to detect ground moving targets embedded in clutter environment

and the estimation of clutter covariance matrix with training samples is required to perform the adaptive processing commonly.In realistic environment

the characteristics of clutter in the cell under test may be different with the training samples in other range cells

which would reduce the precision of the estimated clutter covariance matrix.To improve the performance of target detection

a training sample selecting method based on polarized classification and power grouping is proposed in this paper

which would improve the precision of the clutter covariance matrix.The simulation results demonstrate the effectiveness of the proposed method.