Automatic Mapping of Landslides by the ResU-Net

Abstract

Massive landslides over large regions can be triggered by heavy rainfalls or major seismic events. Mapping regional landslides quickly is important for disaster mitigation. In recent years, deep learning methods have been successfully applied in many fields, including landslide automatic identification. In this work, we proposed a deep learning approach, the ResU-Net, to map regional landslides automatically. This method and a baseline model (U-Net) were collectively tested in Tianshui city, Gansu province, where a heavy rainfall triggered more than 10,000 landslides in July 2013. All models were performed on a 3-band (near infrared, red, and green) GeoEye-1 image with a spatial resolution of 0.5 m. At such a fine spatial resolution, the study area is spatially heterogeneous. The tested study area is 128 km2, 80% of which was used to train models and the remaining 20% was used to validate accuracy of the models. This proposed ResU-Net achieved higher accuracy than the baseline U-Net model in this mountain region, where F1 improved by 0.09. Compared with the U-Net model, this proposed model (ResU-Net) performs better in discriminating landslides from bare floodplains along river valleys and unplanted terraces. By incorporating environmental information, this ResU-Net may also be applied to other landslide mapping, such as landslide susceptibility and hazard assessment.

Constructing a Large-scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates

“…Moreover, the accuracy of these methods can be easily affected by noise and outliers in the image [19]. Two approaches are commonly used in the literature [20]: (1) pixel-based methods, which use spectral information to detect pixels from remote sensing images that correspond to landslides (e.g., [21]- [24]); and (2) objectbased image analysis, which uses both spectral and spatial information for landslide identification (e.g., [18], [25]). These two approaches are commonly integrated with image change detection to identify landslides from multi-temporal remote sensing images.…”

“…Deep learning, a branch of ML, has also been used for landslide mapping [20], [29]- [34]. Deep learning techniques are more efficient in terms of automatic feature engineering directly from satellite imagery.…”

Constructing a Large-scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates

“…Due to the sparse availability of high spatial resolution DEMs for automatic landslide detection, especially in the case of some isolated mountainous regions with complex topography, some researchers use only spectral bands of satellite images to identify regional landslides with acceptable accuracy. [20] proposed a deep network called ResU-Net for landslide identification. The ResU-Net model uses residual blocks [6] in the U-Net architecture.…”

“…The ResU-Net model uses residual blocks [6] in the U-Net architecture. The authors claim that the residual learning used in the encoding path of the U-Net model can address data sparsity problems [20]. Given the success of the U-Net architecture in semantic segmentation, it has been used by multiple researchers for landslide mapping.…”

Semantic Segmentation Deep Learning for Extracting Surface Mine Extents from Historic Topographic Maps

“…UNet and other semantic segmentation methods have been applied to a variety of feature extraction and classification problems and have also been applied to a variety of geospatial and remotely sensed data. For example, modifications of UNet have been applied to the mapping of general land cover change [62], coastal wetlands [63], palm trees [64], cloud and cloud shadows [65], urban buildings and change detection [66][67][68], roads [69], and landslides [70]. Generally, UNet and other FCNs have shown great promise due to their ability to model complex spatial patterns and context while generating data abstractions that generalize well to new data [54,55].…”