Abstract
This article presents a novel dual-path full convolutional network (DP-FCN) model for constructing a landslide inventory map (LIM) with bitemporal very high-resolution (VHR) remote sensing images. Unlike traditional methods for drawing LIM, the proposed DP-FCN directly draws LIMs from the bitemporal aerial images with VHR through a trained deep neural network without generating the change magnitude map. Thus, the proposed approach can effectively reduce the effects of pseudo changes caused by phenological differences rather than landslide events. The proposed DP-FCN model contains two modules, namely, deep feature extraction, and joint feature learning networks. Deep feature extraction aims to reduce redundancy while extracting the high-level deep features from bitemporal images. Joint feature learning establishes the relationship between the deep features of bitemporal images and the ground reference map. Experiments on the real datasets of the landslide sites in Lantau Island of Hong Kong, China, demonstrate the feasibility and superiority of the proposed approach in drawing LIM with VHR remote sensing images. Moreover, compared with the results obtained by the state-of-the-art algorithms, the proposed DP-FCN method achieves the best performance in terms of accuracy for landslide inventory mapping.
Highlights
L ANDSLIDES cause serious economic losses and many casualties annually
From the viewpoint of application, the majority of existing methods have focused on land cover change detection (LCCD), and only limited methods with deep learning techniques have been designed for landslide inventory map (LIM) with very high-resolution (VHR) remote sensing images
These figures demonstrate that false detected (FD) and missing detected (MD) pixels emerged in all the achieved LIMs because despite these aerial images with VHR perform the better visual performance, they are insufficient in spectral bands
Summary
L ANDSLIDES cause serious economic losses and many casualties annually. For example, at least two billion US dollars were expended in landslides in the United States each year [1], [2]. Many methods based on deep learning techniques have been developed for detecting land cover changes with VHR remote sensing images, filters or other pre-operations in these deep learning techniques are usually required for processing raw bitemporal images for these commonly used networks. CMIs are first generated on the basis of the explored high-level features, and changes are indirectly detected from the CMIs. from the viewpoint of application, the majority of existing methods have focused on LCCD, and only limited methods with deep learning techniques have been designed for LIM with VHR remote sensing images.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
