The visualization of the spatial distribution and topological structure of soil pores plays an important role in understanding soil ecological processes. Thus the research on the soil pore skeleton extraction algorithm was carried out. Based on the custom rule structure and the soil pore structure, the performance of the thinning method and distance transformation method to construct the pore skeleton model were evaluated. The experimental results showed that the skeletal model construction effects of the two algorithms were not affected by the model type. Among them, the skeleton model constructed by the distance transformation method missed voxel points and some voxel points tended to be off the center. While the pore skeleton model extracted by the thinning method had perfect centrality, connectivity and topological invariance, whose average skeleton offset distance (0.10mm) was 50% lower than that of the distance transform method (0.15mm). Through the comprehensive analysis of three indicators of refinement, connectivity and centrality, the thinning method had superior ability to describe the shape and topological characteristics of soil pores. The research laid a technical foundation for the study of soil physical structure and hydrological characteristics, and a theoretical foundation for understanding soil function from pore scale.