基于U-Net网络的秭归县建筑物影像识别与空间化  

Image Recognition and Spatialization of Buildings in Zigui County Based on U-Net Network

作　　者：张萍 李垠[1,2] 吕筱 张亦梅 特木其勒[1,2] ZHANG Ping;LI Yin;LYU Xiao;ZHANG Yimei;TEMU Qile(Key Laboratory of Earthquake Geodesy,Institute of Seismology,China Earthquake Administration,Wuhan 430071,China;Hubei Earthquake Agency,Wuhan 430071,China)

机构地区：[1]中国地震局地震研究所地震大地测量重点实验室,武汉430071 [2]湖北省地震局,武汉430071

出　　处：《华南地震》2024年第4期33-39,共7页South China Journal of Seismology

基　　金：中国地震局地震研究所基本科研业务费专项资助项目和中国地震局地壳应力研究所基本科研业务费专项资助项目(306337-12);湖北省地震局基础科研基金项目(2022HBJJ012)联合资助。

摘　　要：基于更新地震应急基础数据库的需求,为快速获取区域建筑物基础信息数据,提出一种建筑物数据空间化方法:利用U-Net全卷积神经网络模型从遥感影像上提取建筑物信息,通过GIS技术将提取的建筑物数据空间网格化,从而获得建筑物空间化格网数据库。以秭归县为研究区域,验证了方法的可行性,所得结果更好地反映了房屋的实际分布情况,为提高地震灾害损失快速评估的精度和准确性奠定基础,为地震应急工作提供更有效的数据支撑。Based on the demand to update the earthquake emergency foundation database,a spatialization method of building data was proposed to quickly obtain the basic information data of buildings within the study area.The building information was extracted from the remote sensing images by using the U-Net full convolutional neural network model,and the GIS technology was used to change the extracted building data into spatial grid data,so as to obtain a spatialized grid database of buildings.Taking Zigui County as the study area,the feasibility of the method was verified.The results obtained can better reflect the actual distribution of buildings,laying a foundation for improving the precision and accuracy of rapid assessment of earthquake disaster losses and providing more effective data support for earthquake emergencies.

关 键 词：U-Net 遥感影像 房屋空间化 地震灾情 快速评估 

分 类 号：P315.9[天文地球—地震学]