Automatic recognition of landslide based on CNN and texture change detection

Landslide is a common disaster that brings huge threat to the safety of human beings' life and property. Because of the remote sensing technology, our efficiency of disaster monitoring has been improved. However, traditional methods usually need the participation of humans thus causing waste of manpower and material resources. A novel method which can conduct automatic recognition of landslide based on convolutional neural network (CNN) and texture change detection is proposed in this paper. Using three steps to gradually narrow the search scope and finally confirm the authentic landslide areas, which outperforms the traditional methods in avoiding a large scale of error detection, efficiency and convenience. The remote sensing image of Shenzhen's landslide areas is chosen to conduct our methods. And the results of the experiments show that the proposed method can effectively extract the areas of landslides and achieve a low commission error.