In order to realize the estimation of soil heavy metal migration in largescale watersheds, the remote sensing information extraction technology plays an important role in such problems. A small and mediumsized watershed (the Daxia river basin) in Northwest China was sampled and investigated. By analyzing the soil and water samples of the Daxia river basin and its tributaries, the contents and distribution characteristics of heavy metals—As, Cd, Cr, Cu and Pb were determined. The contents of five heavy metals were ranked in descending order as the follows: Cr, As, Cu, Pb and Cd. According to the fractal characteristics of river network in river basins, a distributed migration model was put forward to estimate the heavy metal migration in the Daxia river basin. The fractal dimension of the Daxia river basin and its tributaries were extracted and corrected by using grid method and GIS technology. The fractal dimension of the water system was 1.0163, which was less than 1.6, and determined the geomorphology of the Daxia river basin in its infancy. The similarities of the watershed was modeled and verified. And based on the similarities between the fractal dimension and the runoff data, the subbasin runoff volume was further studied to estimate the distributed heavy metal migration, which was verified by comparing the estimated amounts of the SCS-CN model. The migration estimation results in heavy metals—As, Cd, Cr, Cu and Pb in the Daxia river basin were 0.6710t/a, 0.2099t/a, 6.2816t/a, 1.7465t/a and 1.8377t/a, respectively.