Erosion gully has caused the fragmentation of cultivated land in the northeast black soil region，and obtaining the general distribution of erosion gully lies before scientific treatment measures. In order to reduce the man-power cost of visual interpretation，an automatic erosion gully extraction workflow based on multi-remote sensing information was verified，and this workflow was expected to provide method reference to quick erosion gully extraction. The northern Bayan County in Harbin City，was selected as the study area. Firstly，high-resolution satellite imagery in summer and autumn，digital elevation model（DEM），and the third national land resource survey data were collected. Secondly，the erosion gullies were obtained via visual interpretation，and 36 driving-identification feature variables were built. Finally，27 advantaged feature variables were selected，then multi-scale segmentation，random forest method（RF） were utilized to automatically extract erosion gullies. Results show that erosion gullies are mainly distributed in rolling-hill region. Erosion gullies and the background could be distinguished by the elevation difference，slope length factor，and distance to waterflow according to their topographical-hydrological characteristics. Comparing to the background，erosion gullies possess some particular texture characteristics，and could be identified by gray-level cooccurrence matrix（GLCM）. The testing result shows that the user’s accuracy is 54.28%，the producer’s accuracy is 72.16%，and Kappa coefficient is 0.61，it is concluded that the general spatial distribution of erosion gullies could be obtained through automatic extraction based on object-oriented classification method.