Search

Efficiently and automatically acquiring information on earthquake damage through remote sensing has posed great challenges because the classical methods of detecting houses damaged by destructive earthquakes are often both time consuming and low in accuracy. A series of deep-learning-based techniques have been developed and recent studies have demonstrated their high intelligence for automatic target extraction for natural and remote sensing images. For the detection of small artificial targets, current studies show that You Only Look Once (YOLO) has a good performance in aerial and Unmanned Aerial Vehicle (UAV) images. However, less work has been conducted on the extraction of damaged houses. In this study, we propose a YOLOv5s-ViT-BiFPN-based neural network for the detection of rural houses. Specifically, to enhance the feature information of damaged houses from the global information of the feature map, we introduce the Vision Transformer into the feature extraction network. Furthermore, regarding the scale differences for damaged houses in UAV images due to the changes in flying height, we apply the Bi-Directional Feature Pyramid Network (BiFPN) for multi-scale feature fusion to aggregate features with different resolutions and test the model. We took the 2021 Yangbi earthquake with a surface wave magnitude (Ms) of 6.4 in Yunan, China, as an example; the results show that the proposed model presents a better performance, with the average precision (AP) being increased by 9.31% and 1.23% compared to YOLOv3 and YOLOv5s, respectively, and a detection speed of 80 FPS, which is 2.96 times faster than YOLOv3. In addition, the transferability test for five other areas showed that the average accuracy was 91.23% and the total processing time was 4 min, while 100 min were needed for professional visual interpreters. The experimental results demonstrate that the YOLOv5s-ViT-BiFPN model can automatically detect damaged rural houses due to destructive earthquakes in UAV images with a good performance in terms of accuracy and timeliness, as well as being robust and transferable. [ABSTRACT FROM AUTHOR]

Effective extraction of disaster information of buildings from remote sensing images is of great importance to supporting disaster relief and casualty reduction. In high-resolution remote sensing images, object-oriented methods present problems such as unsatisfactory image segmentation and difficult feature selection, which makes it difficult to quickly assess the damage sustained by groups of buildings. In this context, this paper proposed an improved Convolution Neural Network (CNN) Inception V3 architecture combining remote sensing images and block vector data to evaluate the damage degree of groups of buildings in post-earthquake remote sensing images. By using CNN, the best features can be automatically selected, solving the problem of difficult feature selection. Moreover, block boundaries can form a meaningful boundary for groups of buildings, which can effectively replace image segmentation and avoid its fragmentary and unsatisfactory results. By adding Separate and Combination layers, our method improves the Inception V3 network for easier processing of large remote sensing images. The method was tested by the classification of damaged groups of buildings in 0.5 m-resolution aerial imagery after the earthquake of Yushu. The test accuracy was 90.07% with a Kappa Coefficient of 0.81, and, compared with the traditional multi-feature machine learning classifier constructed by artificial feature extraction, this represented an improvement of 18% in accuracy. Our results showed that this improved method could effectively extract the damage degree of groups of buildings in each block in post-earthquake remote sensing images. [ABSTRACT FROM AUTHOR]