Automatic Prediction of Landslides Over InSAR Techniques and Differential Detection Using High-Resolution Remote Sensing Images: Application to Jinsha River

摘要

With a rapidly increasing population on or near steep terrain in southwest of China, landslides have become one of the most significant natural hazards. Thus, quick detection, prediction, and identification of early signals of landslide occurrence are critical for prompt emergency information, rescue efforts, and mitigation of further damage such as collapse of a landslide dam. Accordingly, in this paper, a prototype of early-stage-landslide detection is introduced. We construct a new automatic approach to extract landslides from remote sensing imagery, both optical and radar for quick prediction and detection before disasters, achieved by combining interferometric synthetic aperture radar (InSAR) with differential detection method using multi-temporal high-resolution optical images. The idea behind the novel approach is to identify potential landslides by typical distribution features, deformation, and tendency of landslides, including the result of expert second opinion. To verify the feasibility of this landslide prediction system, a case study was performed in Jinsha River area. It is found that the use of automatic detection of the potential landslides over satellite imagery allows the identification and characterization of the affected areas based on landslide features. It is expected that the prototype of landslide prediction can provide early recognitions before severe landslides occur. Total of eight active small-scaled landslides in early edge covering study region are automatically detected.