Accurate landslide extraction is significant for landslide disaster prevention and control. Remote sensing images have been widely used in landslide investigation, and landslide extraction methods based on deep learning combined with remote sensing images (such as U-Net) have received a lot of attention. However, because of the variable shape and texture features of landslides in remote sensing images, the rich spectral features, and the complexity of their surrounding features, landslide extraction using U-Net can lead to problems such as false detection and missed detection. Therefore, this study introduces the channel attention mechanism called the squeeze-and-excitation network (SENet) in the feature fusion part of U-Net; the study also constructs an attention U-Net landside extraction model combining SENet and U-Net, and uses Sentinel-2A remote sensing images for model training and validation. The extraction results are evaluated through different evaluation metrics and compared with those of two models: U-Net and U-Net Backbone (U-Net Without Skip Connection). The results show that proposed the model can effectively extract landslides based on Sentinel-2A remote sensing images with an F1 value of 87.94%, which is about 2% and 3% higher than U-Net and U-Net Backbone, respectively, with less false detection and more accurate extraction results.

Remote sensing monitoring of glacial lakes is an indispensable tool for identifying and preventing glacial lake disasters. At present, the existing extraction methods of glacial lakes based on Landsat remote sensing image have achieved remarkable results, but the algorithms used lack the ability to analyze glacial lake spectral and shape and texture features, and require manual design parameters to fine tune the automation of the algorithm. As a result, it cannot mine the depth features of glacier lakes in remote sensing images accurately enough. To address these challenges, this study designed a self-attention mechanism module U-net network that enhances the propagation of features, reduces information loss, strengthens the weight of glacial lake areas, restrains the weight of irrelevant features, reduces the influence of low image contrast on the model, and deals with the variety of pixel categories in glacial lakes. These features improve the performance of the model. Based on Landsat-8 images, we first extracted glacial lakes in large-scale alpine areas using a U-net network model. To make it a self-attention U-net network, we introduced the attention mechanism into the step connection part of the U-net network to adjust feature weight, focus on learning glacial lake features, and strengthen the network to extract the glacial lake features. Finally, we selected the combination of bands 3, 5, and 6 and all bands of Landsat-8 images sing the self-attention U-net network to extract glacial lakes in the study area and compared and analyzed the extraction results. The experimental results and analyses revealed that the proposed method can effectively segment glacial lakes from Landsat-8 remote sensing images. Its effectiveness was proven by different evaluation indices. Compared with a standard U-net network, the true positive for the combination of 3, 5, and 6 bands increased by 15.95% and for all bands by 5.79%. The area under curve for the whole study area reached 85.03% for all bands. The improved U-net network can, thus, meet the real time needs of glacial lake disaster information acquisition.

Surface deformation poses a great threat to the safety of Jinchuan mining area production activities. At present, the spatio-temporal evolution law and mechanism of surface deformation in the Jinchuan mining area are unclear, and it is difficult to obtain reliable prediction results using the existing spatio-temporal prediction methods due to the lack of model parameters or relevant data. To solve these problems, this study proposes a new unified convolutional neural network with peephole long short-term memory (CNN-PhLSTM). Small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) technology was used to obtain the spatio-temporal evolution characteristics of surface deformation in the period of 2014–2021. Time series InSAR deformation data are merged into a unified network model in series with a time-distributed CNN segmentation and stacked PhLSTM. The InSAR measurement results are shown to be reliable by comparison and verification with the benchmark and InSAR results of different orbits. The proposed CNN-PhLSTM model was evaluated by mean absolute error and structural similarity (SSIM) evaluation indexes, and was compared with support vector regression (SVR), multilayer perceptron (MLP) and CNN-LSTM models. The results show three continuous subsidence areas, namely the Longshou, second western and third eastern mining areas. The cumulative surface deformation continued to increase from 2014 to 2021. Faults and lithology control the spatial distribution of surface deformation in the Jinchuan mining area. The prediction results demonstrate that the surface deformation range will continue to expand and that time-series surface deformation will show a slow deceleration trend in the next two years.