Due to the opacity of soil, it is difficult to observe the fluid movement and ecological processes inside soil pores in detail, which has always been a difficulty and hot spot in research. Microfluidic chips developed in recent years provide a visualization platform for pore-scale observation. In order to depict a more realistic soil pore structure on the microfluidic chip, this study constructed two two-dimensional soil pore structure models based on the Monte Carlo algorithm and the watershed algorithm. By performing seepage simulation within the structure using the lattice Boltzmann method, the permeability and two-phase flow saturation of the structure were analyzed. Research results show that the pore structure generated by the Monte Carlo algorithm is closer to the real soil sample in terms of pore parameters and permeability, while the pore structure generated by the watershed algorithm is closer to the soil sample in terms of saturation. In addition, the goodness of fit of the multiple linear regression analysis between the simulation results of permeability and saturation and the pore parameters is 0.96 and 0.99 respectively. The hydraulic characteristics of the model are highly correlated with the pore parameters. Therefore, the soil microfluidic structure extraction scheme proposed in this article is better than the traditional cylindrical scheme and is more in line with the actual soil structural characteristics.