The quantitative analysis of soil microstructure is a quite significant content in soil mechanics. The current analytical methods still suffer from several shortcoming, inducing low computational accuracy, poor visualization effects, and limited analysis content. In view of this, based on digital image processing technology, this paper proposes a 3D visual modeling and digital analysis method based on SEM images. Through secondary development on the Grasshopper platform, the digital quantitative processing of soil microstructure is realized. The calculation methods of soil particle surface roughness, surface concavity and convexity, 3D void ratio, 3D pore fractal dimension, and other micro parameters are given. Based on the above methods, the qualitative and quantitative analysis of the microstructure characteristics of Q₂ loess are carried out. The results show that: Q₂ loess is mainly of agglomerate and flocculent structure, and the soil particles are relatively compact and firm. The surface roughness of soil particles is mainly distributed between 0~ 30, and the concavity and convexity are mainly distributed between -40~ 40. The 3D porosity ratio is 0.60~ 0.72, and the 3D pore fractal dimension is 2.52~ 2.63.