Home

The significant increase in computational power has led to an explosion of artificial intelligence applications, in an enormous number of sciences, especially in the field of geosciences. Machine learning, deep learning and other statistical and non-statistical methods are already applied for landslide studies. Even so, our experiment goes beyond these approaches and tries to identify ways that landslides can be mapped at very detailed scales with a minimum effort. The degree of novelty resides in our aim to identify not just the landslide as an object, but furthermore, to map its features into distinct categories. The way we plan to achieve this is by using advanced CNNs (deep learning algorithms) for training and detection of main landslide features. Moreover, we plan to fusion data obtained from optical and thermal cameras to further increase the detection accuracies. Being able to detect landslide features from videos and high-resolution aerial imagery, obtained from UAVs, plays an essential role in rapid response for landslides events management. Besides the feature’s detection, we plan to develop of a risk mapping tool for rapid identification of the exposed human activities in areas with known landslide events.