On the Detection of Possible Landslides in Post-Event Satellite Images: A Probability Map Approach

Abstract

Landslides are a significant hazard to property and livelihoods, causing millions of dollars worth of damage annually throughout the world, but especially in tropical regions such as Malaysia. Automated or semi-automated detection of landslides from aerial or satellite imagery and generating landslide susceptibility or hazard map are two of the main research topics in landslide research. In this paper, we propose a probability map approach in detecting possible landslide regions from satellite or aerial images. The detected landslides, tabulated as landslide inventory map, will be useful as the ground truth for evaluating landslide susceptibility map, or even used as one of the causative factors for the susceptibility map itself. The proposed probability map is computed using only colour information, but demonstrated very promising performance in locating potential landslide regions; thus provides a strong platform to locate actual landslides by incorporating texture and shape features in the future.