震后建筑物受灾程度遥感监测方法研究*
杨玉荣1,张仕廉2,铁中用1,周 超1

(1.四川工程职业技术学院 建筑工程系,四川 德阳 618000; 2.重庆大学 建设管理与房地产学院,重庆 400045)

震后建筑物; 受灾程度; 多光谱遥感图像; 纹理特征; 支持向量机; 玉树地震

Analysis of Remote Sensing Monitoring Methods for Postseismic Damage Level of Buildings
YANG Yurong1,ZHANG Shilian2,TIE Zhongyong1,ZHOU Chao1

(1. Department of Constructional Engineering,Sichuan Engineering Technical College,Deyang 618000,Sichuan,China)(2. School of Construction Management and Real Estate,Chongqing University,Chongqing 400045,China)

post-earthquake buildings; level of damage; multispectral remote sensing images; texture features; support vector machine

备注

基于传统Fisher-SVM的震后建筑物受灾程度遥感监测方法未考虑最优分类超平面的限制条件,图像分类结果精度低,建筑物受灾程度监测不准确,提出面向对象的震后建筑物受灾程度遥感监测方法。其基本原理为:采用面向对象的多光谱遥感图像分割方法获取震后建筑物分割对象,再依据纹理特征的对比度、差异性以及方差这3种参数在震前与震后建筑物中的分布规律,提取震后建筑物分割对象的纹理特征,最后采用支持向量机分类方法,将震后建筑分割图像纹理特征输入到支持向量机分类学习,经过训练输出最优的震后建筑物受灾纹理特征分类结果,完成震后建筑物受灾程度分析。利用该方法对2010年青海玉树7.1级地震后建筑物受灾状况进行分析,结果表明:通过优化调整支持向量机最优分离超平面算式,可以对特征纹理进行较好分类,分析结果与实地考察结果基本一致。

The traditional Fisher-SVM remote sensing method for post-earthquake buildings damage degree monitoring does not consider the constraints of the optimal classification hyperplane. And the accuracy of the obtained classification results is low,the post-earthquake monitoring of buildings is inaccurate,therefore,we proposed an object-oriented post-earthquake remote sensing monitoring method for disaster severity of buildings. First,the basic principle of this method is as follows:it uses object-oriented multispectral remote sensing image segmentation method to obtain post-earthquake building segmentation objects. Second,it analyzes the distributing disciplinarian of three parameters(contrast,dissimilarity,and variance)in pre-and post-earthquake buildings to extract the texture features of the segmented objects after earthquake. Finally,it learns to classify by inputting the texture features of the post-earthquake building segmentation image into support vector machine,then through training,output the best classification results of texture feature to achieve the remote sensing detection for damage degree of building after earthquake. This method is used to analyze the disaster situation of buildings in 2010 Qinghai Yushu M7.1 earthquake. The results of the experiment confirm that the proposed method can effectively achieve the remote sensing monitoring for damage degree of building after an earthquake.