Research on Landslide Detection in SAR Images Based on Multi-channel Change Detection Fusion

Abstract

Abstract The landslides have resulted in the grave loss of human life and economic damage. Currently, emergency investigations predominantly employ manual field investigations and remote sensing interpretation. Traditional methods for assessing landslides are time-consuming and ineffective for widespread and numerous landslides. In response, many scholars have successfully employed optical remote sensing imagery to extract landslides. Nonetheless, landslide regions are typically accompanied by copious rainfall along with inclement weather conditions, such as cloudy or overcast skies, which limits the observation capabilities of optical remote sensing methods. At this juncture, the advantages of SAR systems that can penetrate clouds, rain, and fog, and work around the clock under any weather conditions are highlighted. This paper proposes a framework using multi-channel SAR images to extract landslides. The framework consists of fused multi-channel change detection images and traditional detection methods, resulting in effective landslide detection. The proposed method was verified in an experiment using landslide data triggered by an earthquake in Milin County, Tibet, on November 17, 2017.