面向对象分类的震区建筑物损害快速评估  

Rapid assessment of earthquake damage to buildings based on object-oriented classification

作　　者：李方方[1] 马超[1] 张桂芳[2] 郭增长[1] 

机构地区：[1]河南理工大学测绘与国土信息工程学院,河南焦作454000 [2]中国地震局地质研究所地震动力学国家重点实验室,北京100029

出　　处：《河南理工大学学报（自然科学版）》2011年第1期55-60,共6页Journal of Henan Polytechnic University（Natural Science）

基　　金：国家杰出青年基金资助项目(50525414);河南理工大学博士基金资助项目(B2009-46)

摘　　要：建筑物倒塌是地震灾害中最为严重的一种破坏形式,直接造成震区生命财产的损失.震区建筑物的破坏监测对灾情速报、灾害等级划分、灾害评估、应急抢险、次生灾害监测和灾后重建规划等具有重要的指导意义.传统的面向像元的分类方法不能够很好地提取出建筑物信息,而面向对象的分类能够融合除像元光谱信息以外的形状、纹理和位置关系等多种信息,分类更加客观准确.利用玉树Ms7.1地震前后航空航天多源遥感数据,采用面向对象的分类方法,提取建筑物的覆盖面积,然后统计出建筑物的倒塌率为58.6%,与影像目视解译倒塌率60.4%的结果吻合较好.The building collapse is the most severe earthquake damage; Earthquake directly caused the loss of life and property, and it is important to quick report of the disaster monitoring, classification, assessment, emergency rescue, secondary disaster monitoring and reconstruction planning etc.. Conventional pixel - orien- ted classification method is not able to extract the building information exactly, but object - oriented classifica- tion can integrate the shape, texture and position information besides spectral information, which is more objective and accurate. In this paper, Yushu Ms7.1 non - homologous aerospace remote sensing data before and after the earthquake is used to extract the building coverage by using object - oriented classification method. The building collapsed rate was expected 58.6% and close to 60.4% obtained from visual interpretation.

关 键 词：面向对象分类 建筑物倒塌率 多源遥感数据 

分 类 号：P315[天文地球—地震学]