The dynamic characterization of the frozen wall development process across the entire region is hindered by the insufficient spatial resolution associated with traditional artificial ground freezing (AGF) temperature field monitoring methods. A non-destructive monitoring method based on Rayleigh wave dispersion analysis was proposed. By exploiting the sensitivity of Rayleigh waves to variations in the wave velocity of the medium, dispersion curves were inverted to derive the evolution of the stratum's wave velocity structure during the freezing process. Consequently, the dynamic quantitative characterization of the frozen wall's spatial development morphology was facilitated. Using the freezing gap project of Zhengzhou Metro Line 7 as a case study, a hydro-thermal coupling model for artificial ground freezing was established, and theoretical reference values for the evolution of the frozen wall thickness were calculated. The applicability of Rayleigh wave technology for monitoring frozen wall development was evaluated by comparing the frozen wall thickness obtained from on-site Rayleigh wave inversion with the simulation results. The research results indicate that the stratum's wave velocity structure is significantly altered by the formation of the frozen wall. At 16 days of freezing, a sudden increase in wave velocity in the 5-15 Hz range is observed in the Rayleigh wave dispersion curve, by which the initial formation of the frozen wall is indicated. At 30 days of freezing, a further increase in wave velocity and the stabilization of dispersion characteristics are observed, suggesting that the frozen wall continues to develop and enters a stable strengthening stage. Furthermore, based on the inversion analysis of Rayleigh wave dispersion curves, the dynamic development thickness of the frozen wall can be estimated. At 16 days and 30 days of freezing, inverted thicknesses of 9.5 m and 12.3 m are obtained, respectively. Compared to the theoretical reference values, relative errors of 5.0% and 2.5% and root mean square error (RMSE) of 12% and 9% are presented, respectively. In comparison with finite element calculations, the majority of samples are covered by the 85% confidence interval of the inversion deviation. An overall average relative error of less than 15% and an average monitoring precision of 0.5 m are achieved, through which continuous monitoring of the frozen wall development process is facilitated. Thus, a high-resolution, non-destructive monitoring method is provided for artificial ground freezing engineering.