Hot-dry Valley of Jinsha River is one of the typical serious degenerated regions, The area has abundant sunshine, rich heat resource, low precipitation but high evaporation which leads to severe soil water deficit during the dry season. The adverse environment resulted in lower biological productivity and difficulty in vegetation. A lot of studies proved that vegetation restoration is an effective approach of ecosystem restoration in the dry-hot valleys of Jinsha River. Using double-rings method ,principal component analysis and Pearson correlation analysis, the characteristics of soil infiltration and water-holding capacity of three kinds of vegetation restoration models(taking bare land for comparison) were studied in Xiaokuashan watershed of Yuanmou Dry-hot Valley, Yunnan Province. The results indicated that: ①Soil saturated water storage, water-holding capacity of non-capillary porosity and water-holding capacity of capillary porosity were all shown as the sequence of Tamarindus indica L. + Paspalum notatum Flugge(1317.0, 118.5, 1198.5 t/hm²) > Leucaena leucocephala (Lam.) de Wit (1152.0, 99.0, 1053.0 t/hm²)> Heteropogon contortus(L.) Beauv. (1108.5, 89.5, 1029.0 t/hm²)> bare land(954.0, 66.0, 888.0 t/hm²). The improvement function played by each model on water-holding capacity of non-capillary porosity was greater than that played by corresponding model on water-holding capacity of capillary porosity. ②Infiltration eigenvalues of three models and bare land were shown as initial infiltration rate(T. indica L.+ P. notatum Flugge: 7mm/min; H. contortus (L.) Beauv.: 8.37mm/min; L. leucocephala (Lam.) de Wit:6.01mm/min; bare land:0.62mm/min)＞average infiltration rate(T. indica L.+ P. notatum Flugge:5.93mm/min; H. contortus (L.) Beauv.: 5.06mm/min; L. leucocephala (Lam.) de Wit:3.94mm/min; bare land:0.53mm/min)＞stable infiltration rate(T. indica L.+ P. notatum Flugge:4.94mm/min; H. contortus (L.) Beauv.: 2.88mm/min; L. leucocephala (Lam.) de Wit:3.29mm/min; bare land:0.27mm/min). Wether in dry season or rainy season, we can integrate estimation about soil permeability values (F) by principal component analysis, soil permeability of different models was presented as T. indica L.+ P. notatum Flugge (F:1.2271)> H. contortus (L.) Beauv. (F:0.5365) > L. leucocephala (Lam.) de Wit (F:0.2688)> bare land (F:-2.0323). ③Pearson correlation analysis showed that permeability of three models and bare land had a significant positive linear correlation with soil total porosity, non-capillary porosity and the organic matter content, and a significant negative linear correlation with soil bulk density. Based on which dominant-factor equations were established. ④General empirical model was the best one to describe the soil infiltration processes of this area.