Landslide Prediction using CNN with Data Modeling and Feature Reduction using Median Pooling
Ruchika Malhotra, Abhinav Thapper, Aditya Kulraj Kunwar, Ashutosh Raturi · 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) · 2021
Landslides are one of the most severe and pernicious natural phenomena which occur on Earth. They particularly occur in sloped regions where the integrity of the topsoil of the land has been compromised. The reasons for this are multifaceted, depending on the lithography of the region in question. A landslide can be triggered by intense inundation, cloudburst, earthquakes, deforestation, which at times can be a consequence of urbanization, amongst other factors. They have a detrimental effect on the geography, local economy of the region they occur in and cause immense monetary loss as well. There is a need for timely detection of landslides for establishing correct preventive measures and technological advancements, especially in the field of deep learning are being used for the same. In this paper we present an approach based on Convolutional Neural Networks, using median pooling for feature reduction applied on a dataset with highly inclusive features like vegetation index, water-based index, temperature wind speed, precipitation etc. The architecture of the models is discussed and the preprocessing of features is done as well. We finish with a discussion on model performance and possible improvements.