Abstract: Surface wave dispersion curve inversion is an important geophysical method for obtaining the velocity and thickness distribution of underground shear wave.Conventional linear inversion methods,such as iterative least squares,relying on the initial model and multiple solution,are easy to fall into local minimum and low inversion accuracy.The stochastic inversion method based on Bayesian theory is a nonlinear inversion method which can integrate prior information.This method does not need initial model,only uses prior information to sample the model randomly,and selects and accepts the appropriate inversion model according to the probability distribution.It achieves the accurate estimation of the detail information.In this paper,the authors present a Bayesian Markov Monte Carlo (MCMC) stochastic inversion method based on GPR data constraints to invert the Rayleigh-waves dispersion curve.In the inversion process,by randomly changing the model parameters and calculating the likelihood function of the dispersion curve and the actual dispersion curve,researchers can choose whether to accept the new model parameters,repeat this process continuously,and finally get the best fitting result with the actual dispersion curve and the posterior probability density distribution of the VS solution.The typical numerical model test and field seismic data demonstrate that,compared with the conventional unconstrained stochastic inversion,the proposed method can effectively reduce the multiple solution and improve the efficiency and accuracy.