Abstract: The microtremor imaging method is widely used in engineering investigation. However, when applied to frozen ground, variations in the ice content of permafrost often lead to the presence of interlayers with abrupt velocity changes, which compromises the accuracy of conventional inversion methods. To address this issue, this study develops a two-step joint inversion approach for microtremor imaging of frozen ground based on Monte Carlo (MC) and Damped Least Squares (DLS) methods. The MC algorithm, with its multi-restart and random walk strategies, is employed to search the global space for models with the minimum misfit error. The resulting model is then further optimized using the DLS method, enabling precise detection of the subsurface velocity structure. The accuracy and reliability of the proposed method are validated through synthetic tests using typical stratigraphic models. The method is then applied to process and image real microtremor data acquired from a permafrost region. The results demonstrate that, under the complex stratigraphic conditions typical of frozen ground, this method can effectively delineate the permafrost table and stratigraphic interfaces. The inversion results show good agreement with actual geological characteristics, providing relevant parameters for route selection and hazard mitigation in cold-region road construction.