Lyu, Wanrong ， Qian, Xiaqing ， Dai, Wen ， Li, Chenrui
2026-12-31 GEOCARTO INTERNATIONAL 2026 41(卷), 1(期), (null页)
Hills are fundamental geomorphological units in Gobi deserts, and accurate delineation of their boundaries is essential for understanding surface evolution and supporting desertification management. This paper provides an automated high-resolution hill boundary extraction method that integrates positive-negative (P-N) terrain segmentation and multi-directional hillshading derived from UAV-based DEMs. The P-N method captures local relief patterns but tends to omit low-relief hill margins, whereas hillshading enhances edge visibility but may fragment ridge-top regions. By fusing the binary outputs of both methods, the integrated method improves boundary completeness and spatial consistency. Experiments in the Santang Lake area, Xinjiang, China, demonstrate that the proposed method achieves an overall accuracy of 0.87 and a kappa coefficient of 0.72, representing an improvement of approximately 15%-20% compared with either method alone. The proposed method offers an efficient and reproducible tool for hill mapping in low-relief Gobi environments and can support geomorphological analysis and desert ecological monitoring.