Abstract: Inversion and interpretation of underground structure are ultimate aim of Ground Penetrating Radar (GPR) working. Most of inversion problems are non-linear, hence, investigations of non-linear inversion methods are significant. In this paper, an improved Particle Swarm Optimization (PSO) is used to solve GPR inverse problem. Comparison results with other inversion including the genetic algorithm show that the proposed method has higher accuracy and better simplicity; inversion results under a condition of complicated model, multi- parameter and low SNR indicate the effectiveness dealing with multi-parameter and better anti-noise ability of the proposed algorithm; the inversion results of actual measurement data further verify the feasibility of this algorithm.