Artificial Intelligence Techniques for Landslide Prediction Using Satellite Imagery

In mountainous regions, landslides can be triggered by various natural factors, including heavy rainfall, earthquakes, and soil moisture, as well as human activities such as unplanned construction. These landslides can result in significant property damage and loss of life, making automatic prediction methods essential for prevention. Recently, machine learning algorithms have been employed to facilitate the automatic identification of landslides. Several feature extraction and classification techniques have been applied to satellite imagery for semiautomatic detection and prediction of landslides. However, there has been limited research focused on achieving fully automatic detection with satisfactory accuracy. One of the primary challenges in classifying and predicting landslides from satellite images is the need for a suitable database for training that can produce highly accurate results. This study aims to conduct a thorough examination of various techniques used for detecting and classifying landslides through satellite imagery. A total of fifty research papers focusing on machine learning and deep learning algorithms from reputable journals have been analyzed. This article presents a summary of the performance of different classification techniques found in recent literature, along with a comparison and discussion regarding their accuracy. Based on the identified gaps, an effective prototype for landslide classification is proposed, featuring a slightly modified version of the deep learning model ResNet101, which achieves an accuracy of 96.88% when tested on an augmented dataset of 770 satellite images from Beijing. Furthermore, this article provides researchers with an updated overview of the current state and potential directions for machine and deep learning algorithms in landslide detection. The techniques discussed will serve as a valuable resource for identifying research gaps, guiding new researchers, and encouraging innovative exploration in the field of landslide classification using satellite imagery.