In order to solve the problem that difficult to extract distribution information of sublateral canal without water or with less water caused by low resolution of remote sensing image, a hierarchical classification method of feature combination was proposed, which was based on object-oriented classification method. Bangleng village in Hetao Irrigation District was chosen as the study region, and multi-spectral images were obtained by using fixed-wing unmanned aerial vehicle (UAV) which carried multi-spectrum camera (520~920nm). After a lot of experiments, finally, the segmentation threshold value of 65 and the combined threshold value of 90 were chosen as the best remote sensing image segmentation parameters, then can interpret the obtained high resolution multi-spectral image data. By comparing the spectrum, geometry, spatial relationships between sublateral canal and the other surface features, different levels of classification rules were established to extract sublateral canal distribution information. And 14 sublateral canals in the study region were extracted. The results showed that due to the strong absorption in near infrared spectrum of water, the extraction accuracy of sublateral canal with water was 97.8%;the extraction accuracy of sublateral canal with less water or no water was 75.7%. Using UAV remote sensing techniques and combination of features object-riented hierarchical classification method provided a new way to identify sublateral canal in irrigation area. And future research should focus on eliminating the effect of trees, weeds and gate, as well as extracting canal which in both sides had surface features with close spectrum.