Abstract: The Loess Plateau in West China covers a wide area and has many landslides. Due to problems with field investigations and visual interpretation, such as the time and effort involved, as well as relatively long cycles of data acquisition and update, the high spatial resolution remote sensing data and digital elevation model (DEM) have been used for regional identification of loess landslides in Baota District, Yan'an City, Shaanxi Province, China. The regional loess landslides have been identified by using the object-oriented classification method, this method is to use the image segmentation combined with spectral-spatial and geomorphological features and based on band selection and scale analysis to analyze the landslides. Analysis of the study site showed that the recognition accuracy for the landslide back scarp and landslide mass is 78.9% and 73.6%, respectively; the landslide back scarp is easier to recognize than the landslide mass. This method is important for landslide cataloging, earth science analysis, and image interpretation.