李金香, 赵朔, 金花, 李亚芳, 郭寅. 2019: 结合纹理和形态学特征的高分遥感影像建筑物震害信息提取. 地震学报, 41(5): 658-670. DOI: 10.11939/jass.20190014
引用本文: 李金香, 赵朔, 金花, 李亚芳, 郭寅. 2019: 结合纹理和形态学特征的高分遥感影像建筑物震害信息提取. 地震学报, 41(5): 658-670. DOI: 10.11939/jass.20190014
Li Jinxiang, Zhao Shuo, Jin Hua, Li Yafang, Guo Yin. 2019: A method of combined texture features and morphology for building seismic damage information extractionbased on GF remote sensing images. Acta Seismologica Sinica, 41(5): 658-670. DOI: 10.11939/jass.20190014
Citation: Li Jinxiang, Zhao Shuo, Jin Hua, Li Yafang, Guo Yin. 2019: A method of combined texture features and morphology for building seismic damage information extractionbased on GF remote sensing images. Acta Seismologica Sinica, 41(5): 658-670. DOI: 10.11939/jass.20190014

结合纹理和形态学特征的高分遥感影像建筑物震害信息提取

A method of combined texture features and morphology for building seismic damage information extractionbased on GF remote sensing images

  • 摘要: 为提高震害信息获取时效性,对基于我国国产高分遥感影像的建筑物震害信息提取方法进行深入研究,本文以2017年5月11日新疆塔县MS5.5地震为例,利用该地震前后极灾区高分遥感影像,利用结合纹理和形态学特征的方法进行了建筑物震害信息提取,通过变化检测分析获取了极灾区建筑物震害信息,并与基于像元级和基于目标级的信息提取结果进行对比,采用震后无人机影像目视解译结果对本文结果进行了精度验证。结果表明:通过缩减研究区范围可大力提高数据提取精度和速度;运用灰度共生矩阵、二值化、数学形态学等方法对影像进行迭代运算,能较好地提取高分遥感影像中的建筑物信息;通过对地震前后建筑物提取结果进行变化检测分析,能够有效地提取完全倒塌的建筑物,信息提取总体精度为90.45%,比基于像元级和基于目标级信息提取结果的精度分别提高了5.78%和5.23%,可为震后快速确定人员压埋点、部署救援力量提供决策依据,提高地震应急救援的时效性。

     

    Abstract: It is of great significance to study the methods in the extraction of building seismic damage information based on high-resolution remote sensing images in China, which can improve the timeliness of seismic damage information acquisition. Taking an earthquake with MS5.5 occurred near Taxkorgan Tajik Autonomous County, Kashi Prefecture, Xinjiang Uygur Autonomous Region, China, on May 11, 2017, as an example, based on high-resolution remote sensing images before and after the earthquake, building information was extracted by the method of combined texture features and morphology. Building damage information in extremely disaster areas was extracted through change detection and analysis, and then compared with the results extracted by pixel-based and object-based methods. Finally, the accuracy was verified by visual interpretation results of unmanned aerial vehicle images after the earthquake. The results show that the accuracy and speed of data extraction can be greatly improved by reducing the scope of the studied area. Using gray level co-occurrence matrix, binarization, mathematical morphology and other methods we can extract building information from GF remote sensing images more effectively. Through the change detection and analysis of building extraction results before and after the earthquake, completely collapsed buildings can be effec-tively extracted. The overall accuracy of information extraction is 90.45%, which is 5.78% and 5.23% higher than that of pixel-based and object-based information extraction, respectively. The completely collapsed buildings information can provide decision-making basis for rapid determination of people buried places and deployment of rescue forces after earthquakes, and improve the timeliness of earthquake emergency rescue.

     

/

返回文章
返回