Abstract: Remote sensing, a technology used for quickly and extensively acquisition of land cover information, provides a reliable data source for complex natural resource survey.Aiming at the problem of mountain boundary recognition, an unsupervised statistical learning method was proposed to extract mountain features using remote sensing satellite images for modeling of mountain features.Specifically, DBSCAN algorithm and edge detection ideas were used to identify the mountain area and extract the mountain boundary.This approach recognizes the mountain boundary automatically, which does not rely on marking the ground truth manually.In the experiment, the Landsat 8 remote sensing satellite image data of Anqing City were used to effectively identify the mountainous area and extract the boundaries of the mountains.Through qualitative and quantitative analysis, the reliability of the proposed method was verified.Moreover, it proved the application potential of remote sensing technology and statistical learning theory in the field of natural resource survey.