Search

Landslides are one of the most common geological disasters in China, characterized by suddenness and uncertainty. Traditional methods are not sufficient for the accurate identification, early warning, and forecasting of landslide disasters. As high-resolution remote sensing satellites and interferometric synthetic aperture radar (InSAR) surface deformation monitoring technology have been leaping forward, the traditional methods of landslide monitoring data sources are limited, and there have been few effective methods to excavate the characteristics of the spatial distribution of landslide hazards and their triggering factors, etc. In this study, an area extending 10 km from the VII isobar of the Gengma earthquake was taken as the study area, and 13 evaluation factors were screened out by integrating the factors of InSAR surface deformation, topography, and geological environment. Landslide susceptibility was evaluated through the Bayesian optimized convolutional neural network (BO-CNN), and the Bayesian optimized random forests (BO-RF) and particle swarm optimization support vector machines (PSO-SVM) models were selected for comparative analyses. The accuracy of the model was evaluated by using three indices, including the ROC curve, the AUC value, and the FR value. Specifically, the ROC curves of PSO-SVM, BO-RF, and BO-CNN were close to the upper-left corner, indicating excellent model performance. Moreover, the AUC values were computed as 0.9388, 0.9529, and 0.9535, respectively, and the FR value of landslides in the high susceptibility area of BO-CNN reached up to 14.9 and exceeded those of PSO-SVM and BO-RF, respectively. Furthermore, the mentioned values of the SVM and BO-RF models were 4.55 and 3.69 higher. The experimental results indicated that, compared with other models, the BO-CNN model used in this study had a better effect on landslide susceptibility evaluation, and the research results are of great significance to the disaster prevention and mitigation measures of local governments. [ABSTRACT FROM AUTHOR]

At present, in the field of video-based human action recognition, deep neural networks are mainly divided into two branches: the 2D convolutional neural network (CNN) and 3D CNN. However, 2D CNN's temporal and spatial feature extraction processes are independent of each other, which means that it is easy to ignore the internal connection, affecting the performance of recognition. Although 3D CNN can extract the temporal and spatial features of the video sequence at the same time, the parameters of the 3D model increase exponentially, resulting in the model being difficult to train and transfer. To solve this problem, this article is based on 3D CNN combined with a residual structure and attention mechanism to improve the existing 3D CNN model, and we propose two types of human action recognition models (the Residual 3D Network (R3D) and Attention Residual 3D Network (AR3D)). Firstly, in this article, we propose a shallow feature extraction module and improve the ordinary 3D residual structure, which reduces the parameters and strengthens the extraction of temporal features. Secondly, we explore the application of the attention mechanism in human action recognition and design a 3D spatio-temporal attention mechanism module to strengthen the extraction of global features of human action. Finally, in order to make full use of the residual structure and attention mechanism, an Attention Residual 3D Network (AR3D) is proposed, and its two fusion strategies and corresponding model structure (AR3D_V1, AR3D_V2) are introduced in detail. Experiments show that the fused structure shows different degrees of performance improvement compared to a single structure. [ABSTRACT FROM AUTHOR]