测绘通报 ›› 2024, Vol. 0 ›› Issue (2): 69-73.doi: 10.13474/j.cnki.11-2246.2024.0212

• 学术研究 • 上一篇    下一篇

面向震后快速灾害评估的遥感影像房屋数据空间化构建方法

张萍1,2, 李必军3, 李垠1,2, 张亦梅1,2, 特木其勒1,2, 刘可1,2, 李治君4   

  1. 1. 中国地震局地震研究所地震大地测量重点实验室, 湖北 武汉 430071;
    2. 湖北省地震局, 湖北 武汉 430071;
    3. 武汉大学测绘遥感信息工程国家重点实验室, 湖北 武汉 430079;
    4. 自然资源部信息中心, 北京 100812
  • 收稿日期:2023-06-21 出版日期:2024-02-25 发布日期:2024-03-12
  • 通讯作者: 李必军。E-mail:lee@whu.edu.cn
  • 作者简介:张萍(1993—),女,硕士,工程师,主要从事遥感与GIS技术应用研究工作。E-mail:zping@whu.edu.cn
  • 基金资助:
    中国地震局地震应急青年重点任务(CEAEDEM202213);中国地震局地震研究所基本科研业务费专项和中国地震局地壳应力研究所基本科研业务费专项(306337-12);湖北省地震局基础科研基金(2022HBJJ012)

Housing data spatialization research based on remote sensing images for rapid loss assessment after earthquakes

ZHANG Ping1,2, LI Bijun3, LI Yin1,2, ZHANG Yimei1,2, Temuqile1,2, LIU Ke1,2, LI Zhijun4   

  1. 1. Key Laboratory of Earthquake Geodesy, Institute of Seismology, China Earthquake Administration, Wuhan 430071, China;
    2. Hubei Earthquake Agency, Wuhan 430071, China;
    3. State Key Laboratory of Information Engineering in Surveying Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China;
    4. Information Center of Ministry of Natural Resources, Beijing 100812, China
  • Received:2023-06-21 Online:2024-02-25 Published:2024-03-12

摘要: 基于卷积神经网络方法可高效提取高分辨率遥感影像上的房屋矢量数据,快速获取房屋数据的空间分布数据,提高地震应急基础数据库的更新能力。本文基于轮廓引导和结构感知的编解码器全卷积神经网络(CGSANet)模型和分区域等尺度网格抽样方法,获取了房屋建筑面积及房屋结构类型空间分布模型,具备了复杂区域背景下的多类型房屋数据空间化技术能力;以黄梅县为研究对象,构建了1 km×1 km的房屋数据空间化数据集,实现了不同结构类型的房屋数据识别与鉴定。构建的房屋数据空间化数据集可为地震应急基础数据库提供数据来源,对于提高房屋数据的精度和时效性具有重要意义。

关键词: 遥感影像, 卷积神经网络, 房屋数据, 空间化, 网格抽样

Abstract: The convolutional neural network method can efficiently extract housing vector data from high-resolution remote sensing images, quickly obtain spatialization data of housing data, and improve the updating ability of earthquake emergency database. Based on the contour-guided and local structure-aware encoder-decoder network(CGSANet) model and the equal scale grid sampling method on the basis of partition, this paper establishes the spatialization model of housing construction area and housing structure types, and achieves spatialization of multi-type housing data in complex regional backgrounds. Taking Huangmei county as the study area, the model of housing data spatialization(1 km×1 km) is constructed, and the ability to identify housing data of different structural types is achieved. The model of housing data spatialization constructed can be used to update the earthquake emergency database, and is of great significance for improving the accuracy and timeliness of housing data.

Key words: remote sensing image, convolutional neural networks, housing data, spatialization, grid sampling

中图分类号: