Mapping of irrigation networks in large-scale irrigation districts is very important for the development of modern water-saving irrigation technology, at the same time, it is also important for allocating water reasonably and supplying water safely. However, due to the low resolution of the remote sensing images currently, it is very difficult to extract the irrigation networks in detail. Based on the slope data generated from digital elevation model (DEM) which was obtained from the unmanned aerial vehicle (UAV) aerial photogrammetry collected in March, 2016, and combined the characteristics of irrigation networks, a revised hough transform (RHT) method was adopted, which realized the fast extraction of irrigation networks. Compared with the object-oriented method, supervised-classification method and the field measurement results, the integrity of RHT method can reach about 85.61%. Errors were mainly concentrated in the non liner cannal part, which existed more sediment deposition (slope was not obvious), and also in the cannal dike when unneceassary parallel or crossed cannal were eliminated. According to the characteristics of irrigation networks, the research result was a beneficial exploration for the study of artificial irrigation network system, which was based on the high-resolution DEMs.