面向对象的房屋震害变化检测方法  

Object-oriented Detection of Earthquake Building Damages

作　　者：龚丽霞[1,2] 李强[2] 张景发[2] 曾琪明[1] 刘明众[2] 李成龙[2] 

机构地区：[1]北京大学遥感与地理信息系统研究所,北京100871 [2]中国地震局地壳应力研究所,北京100085

出　　处：《地震》2013年第2期109-114,共6页Earthquake

基　　金：国家"863"计划(2012AA121304);高分计划(E0205/1112/9)资助

摘　　要：遥感图像变化检测是震害评估较为有效的方法,随着卫星影像分辨率不断提高,地物对象正逐渐取代单个像元,成为变化检测的基本单元。本文阐述了面向对象的变化检测方法的原理和数据处理流程,以玉树地震前后的高分辨率遥感影像为例,展示了这一方法提取房屋震害的能力。与直接基于像元的变化检测方法、基于主成分分析的变化检测方法相比较,结果显示面向对象的变化检测方法精度最高,提取房屋变化信息精度达到86%,而后面两种方法的精度分别为59%和68%。Change detection from remote sensing image is an effective method to estimate earthquake damage. With the improvement of of spaceborne image resolution, single pixel is gradually replaced by object as the basic unit for change detection. The principle of ob- ject-oriented change detection method is described in this paper. And then data processing flow is stated. Taking high-resolution remote sensing images pre- and post- the 2010 Yushu earthquake as an example, the accuracy of this method to extract building damage is shown. This method is compared with direct change detection method based on pixel, and based on principal component analysis. The results show object-oriented change detection method has the highest accuracy in building damage estimation.

关 键 词：遥感 面向对象 变化检测 房屋震害 

分 类 号：P315.9[天文地球—地震学]