A Supervised Machine Learning Spatial Tool for detecting terrain deformation induced by landslide phenomena.

Περίληψη

The present study uses Google Earth's imagery, a library of high resolution satellite imagery and aerial photography of the entire Earth's surface to extract spectral and morphometric information in order to detect large terrain deformation and produce a landslide inventory map. The developed supervised machine learning spatial tool for landslide detection was verified by experimentation on a Google Earth Images sequence (16 images) that covered a watershed area with numerous recorded landslide incidences in the municipality of Kimi, Euboea Island, Greece. The evaluation of the outcomes of the performed classification showed satisfactory results, with the overall accuracy presented to exceed 88%.