The registration of multi-source Digital Elevation Models (DEMs) is essential for planetary exploration, as it enhances the spatial and temporal coverage of DEM datasets, enabling a broad range of surveying and mapping applications. However, various factors, including the diversity of terrain representations, the weak texture in planetary environments and minimal topographic relief, undermine the effectiveness of common DEM registration methods, making the alignment of planetary multi-source DEMs still a challenging task. To address the nonlinear expression problem in multi-source DEMs, we compared the matching results of various multi-modal matching methods on DEMs from Mars and moon, and selected the best one to construct a robust coarse-to-fine PMS-DEM registration method. Experiments were conducted using multi-source DEM data on Mars and the Moon to assess the effectiveness of the proposed method. The results of image feature matching experiments demonstrate that the Weighted Structure Saliency Feature (WSSF) method outperforms other existing state-of-the-art multimodal matching methods. Besides, the results of DEM registration experiments demonstrate that the reliability of the proposed method compared with other commonly used registration methods.