Citation

Ghorbanzadeh, O., Meena, S.R., Blaschke, T., Aryal, J. (2019). UAV-Based Slope Failure Detection Using Deep-Learning Convolutional Neural Networks. Remote Sensing, 11(17): 2046. Link to paper

Abstract

This paper analyzes the potential of deep learning convolutional neural networks (CNNs) for slope failure detection using optical data from unmanned aerial vehicles (UAVs) in an area along a road section in the northern Himalayas, India. The study tested different CNN architectures across two separate study areas to evaluate their effectiveness in identifying slope failures from high-resolution UAV imagery. The CNN approach that performed best achieved almost 90% precision, an F-score of 85%, and mean Intersection over Union (mIOU) of 74% using a window size of 64×64 pixels for sample patches and including slope data as an additional input layer. The results demonstrate that incorporating topographic information such as slope significantly improves detection accuracy compared to using optical data alone. This research shows the effectiveness of combining UAV-based high-resolution imagery with deep learning for rapid post-disaster slope failure assessment, offering a practical approach for emergency response and hazard management in mountainous regions where traditional mapping methods may be time-consuming or inaccessible.

AI Visual Interpretation (Beta)