Automatic Detection of Landslides Based on Machine Learning Framework
Devara Meghanadh, Vipin Kumar Maurya, Manish Kumar, R. S. Dwivedi
Abstract
Advancement in the satellite imaging techniques has produced wider applications in disaster reduction and management. Considering the crucial role of rapid detection of landslides in disaster management, this research paper exploits the capability of sentinel-1A imagery in landslide detection. In image processing and feature detection techniques, machine learning based classification methods have gained popularity by proving its potential in mapping of susceptible landslides. Therefore, the primary objective of this research paper is automatic identification of landslides by using machine learning based classifier Random Forest (RF). To accomplish the objective, two Sentinel-1A GRD images, pre and post-event, covering Srinagar-Rudraprayag region, India has been processed with DEM derived products slope, aspect and elevation for landslide detection along the Alaknanda River and National Highway-58. For validation of the results, Google Earth Historic Imagery has been utilized. From the observation of results, RF method confirms its potential by achieving 95% in accurate detection of landslides.