Identification of Landslide Vulnerable Zone in Western Ghats using Various Image Processing Techniques and Machine Learning Techniques

Resumen

This proposed work provides remote sensing approaches for determining landslide zones in
the Western Ghats, to estimate landslide vulnerability ranging in size from large to small. A
landslide weakness zone should be identified utilizing landslide possibility points from a
wide range of sources within the study area. To improve the interpretation of remote sensing
images, a variety of image processing technologies have been used. It uses satellite photos to
extract as much information as possible. The application and user-dependent selection of
certain methodologies or algorithms for a specific research issue is application and userdependent.
Remote sensing is a type of data collection in which data about an object is
gathered without having to interact with it. To effectively interpret the Ground atmosphere,
remotely sensed data must be pre-processed to enhance and realign pictures. Space
exploration, satellite photography, geographic information systems (GIS), agricultural
production tracking, disaster risk management, and a variety of other study fields all rely on
satellite image processing. Satellite photos are digitally collected and then processed to
extract data by machines. The landslide parameter features are then categorized into two
parts: 70% for training sets and 30% for validation. To make a comparison of the
performance of various machine learning algorithms such as Naive Bayes, Multilayer
Perceptron, and stochastic gradient descent modeling for estimation of landslide possibilities
in the study area. The spatial prediction uses various machine learning algorithms such as
Naive Bayes, Multilayer Perceptron, and stochastic gradient descent modeling. By evaluating
the landslide triggering elements, weighting them, and optimizing the architectural
characteristics, the algorithm will estimate the risk assessment of the location in a low,
moderate, or high vulnerability zone. The Naive Bayes classifier delivers the most
trustworthy results, with an overall accuracy rate of 98.34 percent. In experiments, Naive
Bayes outperforms other algorithms.