Abstract:Surface features are closely related to natural disasters and have a significant impact on maintaining the ecological environment and deeply understanding the evolutionary process of the earth’s surface and geological structural characteristics. High spatial resolution digital models constructed through unmanned aerial vehicle (UAV) aerial surveys and structure from motion (SfM) technology were used to analyze the distribution relationship between land cover information and terrain features in the circular tectonic landforms of central Yunnan. The results showed that in areas with a mix of bare rock, bare soil, and vegetation, DeepLabv3+ algorithm was found to have better extraction effectson land cover information in the experimental area compared with the RF algorithm, both qualitatively and quantitatively. Ground points obtained from filtered point clouds, and the Kriging algorithm with the smallest mean error and root mean square error in cross-validation was chosen to construct a 0.1m resolution digital elevation model (DEM) to interpret various terrain factors such as firstorder slopes, second-order slopes, and compound slopes. Based on correlation analysis, six types of terrain factors were selected to construct a comprehensive terrain analysis model (CTAM). After analyzing the connection between the land cover information with the largest coverage area (bare soil and vegetation) and the terrain, within CTAM, the percentage of each grade’s pixel number to the total pixel number was the highest at 28.87% for Grade Ⅱ, with Grades Ⅰ, Ⅳ, and Ⅴ accounting for 18.39%, 13.82%, and 17.29%, respectively. UAV-SfM technology can effectively capture the surface features of circular structures and provide technical means and scientific basis for geological research and resource management in the region.
