Landslides Detection using Aerial Ortho-Images and LiDAR Data

摘要

Traditionally the landslide areas were manually measured from aerial stereo-pair by visual interpretation or on-site surveying. Since those two approaches are tedious and cost, researches focus on automatic method for efficient and effective detection for landslides area. For the purpose of automatic landslide detection, its surface and spectral characteristics should be analyzed at first. In the paper, base on a pre-compiled landslide map, we utilize the airborne LiDAR and color aerial ortho-images to analyze the terrain slope, canopy texture, spectral greenness for the landslides. Experimental results, indicate that the greenness is an important index to discriminate the bare soil from vegetation, it is better to use a radiometric calibrated near-infrared image for detecting non-vegetation area automatically. In this preliminary study, the proposed logical intersection method can get a correct detection rate up to 86%, but a large commission error will be introduced. A further modification of the thresholds considering the landslides that are covered by trees, a more reasonable result could be achieved.