In order to suppress the strong direct wave in the original image of GPR, highlight the weak target signal and realize the accurate location of the hidden danger, a Stolt migration method combining with robust principal component analysis (RPCA) is proposed.Firstly, the synthetic data are generated by using the GPR analysis program package matGPR3.1 to simulate the dam-void model, and the original image of the GPR is obtained through scanning processing.Then the RPCA method is used to extract the hyperbolic signal of the target.Finally, the target signal is focused by Stolt migration technique, and the optimal migration effect is obtained by selecting the appropriate migration velocity through minimum entropy velocity estimation.The experimental results show that the proposed method can effectively suppress the clutter and identify the location of the target void more clearly.Compared with the previous methods, the image obtained by the method has higher resolution and better performance, and has a certain application value in the detection of hidden dangers in water conservancy project.