返回
Deep Learning Image-Based Classification for Post-Earthquake Damage Level Prediction Using UAVs
DOI:10.3390/s25175406.png)
摘要
En 中文
无人机(UAVs)与轻量级深度学习模型的集成代表了基于图像的快速地震后灾损评估的有效解决方案。配备相机的无人机能够捕捉灾区的高分辨率航拍影像,为结构损毁评估提供关键数据。当与轻量级卷积神经网络(CNN)模型结合时,这些无人机可对捕获的图像进行机载处理,实现实时、准确的损毁等级预测,这可能有助于高效引导搜索与救援(SAR)团队的工作方向。本研究探讨了使用MobileNetV3-Small轻量级CNN模型,基于无人机捕获的影像进行实时地震后损毁等级预测的应用。该模型被训练用于分类三个地震后损毁等级,范围从无损坏到严重损坏。实验结果显示,改进后的MobileNetV3-Small模型实现了最低的FLOPs(浮点运算次数),相较于ShuffleNetv2模型降低了58.8%。对最后五层进行微调后,FLOPs仅略微增加约0.2%,但显著提升了准确性和鲁棒性,相较于基准模型性能提升了4.5%。该模型在合并了三个地震后损毁等级数据集的融合数据集上取得了0.93的加权平均F分数。该模型成功部署并在Raspberry Pi 5上进行测试,证明了其在边缘设备应用中的可行性。此次部署突显了该模型在资源受限环境下的高效性与实时性能。
Keyword:
UAVs
lightweight deep learning
post-earthquake damage assessment
MobileNetV3-Small
real-time prediction
期刊
IF:
3.5
论文数:
7.2W
被引数:
20.9W
机构
暂无机构信息
引用论文
OpenVINO Deep Learning Workbench: A Platform for Model Optimization, Analysis and DeploymentOpenVINO深度学习工作台:一个用于模型优化、分析和部署的平台
Zhang, J.; Zhao, H.; Yao, A.; Chen, Y.; Zhang, L.; Liao, H. Efficient Semantic Scene Completion Network with Spatial Group Convolution. arXiv 2019, arXiv:1907.05091. [Google Scholar] [CrossRef]张, J.; 赵H.; 姚A.; 陈Y.; 张L.; 廖H. 高效的空间分组卷积语义场景补全网络. arXiv 2019, arXiv:1907.05091. [Google Scholar] [CrossRef]
Tang, S.; Chen, Z. Understanding Natural Disaster Scenes from Mobile Images Using Deep Learning. Appl. Sci. 2021, 11, 3952. [Google Scholar] [CrossRef]Tang, S.; Chen, Z. 利用深度学习从移动图像中理解自然灾害场景。Appl. Sci. 2021, 11, 3952. [Google Scholar] [CrossRef]
Detection of damaged buildings after an earthquake with convolutional neural networks in conjunction with image segmentation
VISUAL COMPUTER
IF2.9
The impact of class imbalance in classification performance metrics based on the binary confusion matrix
PATTERN RECOGNITION
IF7.6
Post-earthquake structural damage assessment, lessons learned, and addressing objections following the 2023 Kahramanmaras, Turkey earthquakes土耳其Kahramanmaras 2023地震后的震后结构损坏评估,经验教训和解决异议
Ruf, B. Fast Dense Depth Estimation from UAV-Borne Aerial Imagery for the Assistance of Emergency Forces; KIT Scientific Publishing: Karlsruhe, Germany, 2023. [Google Scholar] [CrossRef]Ruf, B. 快速密集深度估计:基于UAV载航空影像的紧急救援力量辅助;KIT科学出版社:德国卡尔斯鲁厄,2023年。 [Google Scholar] [CrossRef]
Shianios, D.; Kolios, P.S.; Kyrkou, C. DiRecNetV2: A Transformer-Enhanced Network for Aerial Disaster Recognition. SN Comput. Sci. 2024, 5, 770. [Google Scholar] [CrossRef]Shianios, D.; Kolios, P.S.; Kyrkou, C. DiRecNetV2: 一种用于航空灾害识别的Transformer增强网络。SN Comput. Sci. 2024, 5, 770. [Google Scholar] [CrossRef]

