A landslide dynamic susceptibility evaluation method and device based on deep learning

Abstract

The application provides a landslide dynamic susceptibility evaluation method and device based on deep learning, relates to the technical field of geological disaster monitoring and evaluation, and comprises collecting time-series SAR data and multi-source static environmental factor data of a sample area, obtaining time-series ground surface deformation factor data through time-series InSAR technology processing; obtaining landslide label data as output labels, and time-series ground surface deformation factors and multi-source static environmental factor data as input labels, to construct a sample data set; constructing a deep learning model, which comprises a time-series feature learning module, a static factor feature learning module and a joint feature learning module, and is respectively used for extracting time-series deformation features, static factor features and performing feature fusion; training the model by using the sample set to obtain an evaluation model; and applying the model to evaluate a target area to generate a landslide susceptibility probability prediction result. The application realizes high-precision evaluation of landslide dynamic susceptibility and overcomes the limitations of traditional evaluation models.

Images