Predicting Landslides and Floods with Deep Learning

Abstract

Landslides are destructive natural disasters that have a strong destructive potential. The field of landslide disaster prevention is built around landslide detection mapping. Land slide detection is one of the fundamental studies in this subject since the goal of landslide analysis is to reduce the likelihood of landslide occurrence by physical intervention. This study summarizes pioneering studies in the area of landslide analysis and presents how to gather and use landslide data for deciliter and cubic centimeter techniques. The most popular analytical indexes for object detection and image segmentation are listed next. This work even suggests a system of algorithms for identifying the presence of floodwater (water hazard) in images taken using cell phones or other optical cameras. Of the several approaches tested, the pretrained VGG-16 network using a logistic regression classifier achieved the best results. Landslide and flood prediction research is conducted by interpreting aerial photographs and conducting field verification. Remote sensing technologies have enabled researchers to use pictures provided by high-resolution laser to identify landslides and flood brought on by significant events.