Landslide detection using probability regression, a case study of Wenchuan, northwest of Chengdu

Landslides have become one of the dominant disasters all around the world. Reliable and efficient landslide mapping is playing a significant role in landslide studies. However, to the best of our knowledge, there is little research on detecting multiple landslides simultaneously from images to stimulate practical cases. In this work, we propose a regression model to detect landslides and investigate its applicability in practical cases. It synthesizes contextual, spectral and geometric features. Among the test images, F-measure of landslide detection has a range between 0.771 and 0.998, validating its robustness and high efficiency.