Abstract: The technique of using Yellow River sediment to fill damaged farmlands in mining subsidence areas not only fully and rationally use Yellow River water and sediment resources, but also effectively increase farmland areas and improve regional environment. Infiltration is an important indicator for farmland soil and water conservation and nutrient retention. Lab infiltration apparatus was used to measure the vertical infiltration of reclaimed farmlands with Yellow River sediment fill. The aim of the study was to obtain the soil infiltration characteristics in Yellow River sediment fill farmland in the second year, and analyze and compare the experimental infiltration results of the first and second years of reclamation. Five infiltration models were selected to fit the characteristic curves of infiltration. The results showed that mean initial infiltration rate and mean stable infiltration rate of control farmland were 13.95 mm·min^-1 and 4.36 mm·min^-1, which were 1.12 times and 2.19 times higher than that of fill farmlands in the second year of reclamation. The discrepancy was caused by the compaction effect of machi- nery used in the land reclamation process. The standard deviations of initial infiltration and stable infiltration rates of fill farmlands were 1.98 mm·min^-1 and 2.1 mm·min^-1, significantly lower than those of control farmland (P < 0.05). Thus the infiltration capacity of fill farmlands was steadier than control land. The mean initial infiltration rate and mean stable infiltration rate of fill farmlands in the second year of reclamation were 12.48 mm·min^-1 and 1.99 mm·min^-1, respectively; 1.51 times and 1.43 times higher than that in the first year of reclamation. This suggested that infiltration rate of fill farmlands increased significantly (P < 0.05) after two years of corn-wheat crop rotation. Thus farming practices, root development and the activities of soil microorganisms and small animals enhanced self-recovery capacity of fill farmlands. Therefore the study proposed cultivation of corps with well-developed root system to improve soil infiltration capacity in fill farmlands. In terms of the estimation of soil infiltration rated, Mezencev model was the best for fill farmland, followed by Kostiakov model, NRCS model, Horton model and the Philip model. For control farmlands, Kostiakov model was the best, followed by Mezencev model, NRCS model, Horton model and Philip model. For both fill farmlands and control farmlands, Mezencev model had the best fit result, followed by Kostiakov model, NRCS model, Horton model and then Philip model. As the difference in assessment index between Mezencev model and Kostiakov model was small (P> 0.05), both models had small fitting errors. However, because some of the parameters of Mezencev model (a semi-empirical model) had real physical meaning, Mezencev model was considered to be better than Kostiakov model. In addition, the rankings of the models based on different assessment indices were basically similar. This led to a considerable consistency in RMSE, R², SSE and Ajust-R²of infiltration models. This study provided research methods for determining farmland infiltration characteristics and the significant results were achieved. The results therefore provided the scientific basis for the reconstruction and improvement of fill farmland soils. The study as well provided certain guiding codes and practical values for the management of farmland irrigation and drainage.