Vegetation restoration can significantly affect soil infiltration process. On this basis, two typical abandoned grasslands, Bothriochloa ischcemum (Linn.) Keng and Artemisia vestita Wall. ex Bess, from theLoess Plateau were studied under six planted densities of 5, 10, 15, 20, 25, and 30 plants/m², and an artificial simulated rainfall experiment was adopted to determine the effect of vegetation characteristics on the soil infiltration process. The result showed that:(1) The increasing of planting density could significantly delay runoff. The initial runoff time ranged from 0.76 s to 5.74 s for Bothriochloa ischcemum (Linn.) Keng grassland and ranged from 0.87 s to 2.08 s for Artemisia vestita Wall. ex Bess grassland under six densities. The initial runoff time increased with vegetation cover, root mass density and organic matter as power functions (R² ≥ 0.18,p<0.05). (2) For Bothriochloa ischcemum (Linn.) Keng grassland, the average infiltration rate, stable infiltration rate and infiltration amount ranged from 0.47 to 0.82 mm/min,0.46 to 0.82mm/min and 7.12 to 11.84 mm respectively; for Artemisia vestita Wall. ex Bess grassland, the infiltration parameters above mentioned ranged from 0.38 to 0.67 mm/min,0.35 to 0.60 mm/min and 5.70 to 10.07 mm respectively. When the planting density was 20 plants/m², the soil infiltration parameters were the maximum.The average infiltration rate, stable infiltration rate, infiltration amount and infiltration coefficient K increased with soil organic matter as power functions (R² ≥ 0.26, p<0.01), and the attenuation coefficient α decreased with the increase of biological crust coverage (p>0.05). (3) The initial runoff time, average infiltration rate, stable infiltration rate, infiltration amount and infiltration coefficient K of Bothriochloa ischcemum (Linn.) Keng grassland were all higher than those of Artemisia vestita Wall. ex Bess grassland, while the attenuation coefficient α was lower than that of the attenuation coefficient. It showed that the Bothriochloaischcemum (Linn.) Keng grassland had high soil infiltration capacity. In general, infiltration amount (A) would be simulated by root mass density (RMD) and soil organic matter (SOC) (A=2.77×RMD^0.149SOC^0.614,R²=0.663, NSE=0.653). The purpose of this study was to provide data sources and theoretical basis for the eco-hydrological process and vegetation construction of the abandoned grassland on the Loess Plateau.