Automated Landslide Detection using Ensemble Learning

Landslides pose significant threats to human life, infrastructure, and the environment, making timely detection crucial for effective disaster management. This research paper introduces a novel approach to landslide detection utilizing Convolutional Neural Networks (CNNs) in conjunction with ensembl...

Full description

Saved in:
Bibliographic Details
Published in2024 International Conference on Electrical Electronics and Computing Technologies (ICEECT) Vol. 1; pp. 1 - 6
Main Authors Jain, Tanishka, Agrawal, Subhash
Format Conference Proceeding
LanguageEnglish
Published IEEE 29.08.2024
Subjects
convolutional neural network (CNN)
Convolutional neural networks
deep learning
Disaster management
Ensemble learning
Feature extraction
graphical user interface (GUI)
Graphical user interfaces
landslide
Landslides
Remote sensing
Robustness
Satellite images
Terrain factors
Online AccessGet full text

Cover

More Information
Summary:Landslides pose significant threats to human life, infrastructure, and the environment, making timely detection crucial for effective disaster management. This research paper introduces a novel approach to landslide detection utilizing Convolutional Neural Networks (CNNs) in conjunction with ensemble learning techniques. Leveraging the inherent capabilities of CNNs to extract discriminative features from remote sensing imagery, multiple CNN models are trained and aggregated into an ensemble. Through extensive experimentation on diverse satellite imagery datasets, the effectiveness of the ensemble model is evaluated against standalone CNNs and traditional methods. Results demonstrate that the ensemble approach outperforms individual CNN models and baseline methods regarding accuracy and robustness across varying environmental conditions, enhancing landslide detection capabilities and contributing to more effective disaster management strategies. Our Model achieves 95% Precision, 91% Recall, and 93% f1-score which makes it better than all comparative methods. After classification, we make predictions through GUI by uploading images, in which we are using the Nepal dataset.
DOI:10.1109/ICEECT61758.2024.10739015