汶川地震次生灾害提取——面向对象影像分类技术的应用  

Extracting secondary disaster of Wenchuan earthquake:application of object-oriented image-classifying technology

作　　者：安立强[1] 张景发[1] 赵福军[2] 

机构地区：[1]中国地震局地壳应力研究所,北京100085 [2]黑龙江科技学院,黑龙江哈尔滨150027

出　　处：《自然灾害学报》2011年第2期160-168,共9页Journal of Natural Disasters

基　　金："十一五"国家科技支撑计划(2006BAC13B03-0102)

摘　　要：遥感技术可快速准确地提取滑坡、崩塌、泥石流、堰塞湖等次生灾害信息。面向对象影像分类技术主要通过多尺度分割,并根据对象的特征进行模糊分类,可以很好的利用影像上下文信息,提高分类精度,与传统方法相比优势明显。利用汶川地震的震前、震后的福卫二号等数据进行研究,使用面向对象方法对北川附近灾情进行了变化检测分析;并结合DEM数据对唐家山堰塞湖璇坪乡区域震后初期的水体面积和蓄水增量进行了分析。在此基础上,总结了提取地震次生灾害的关键技术和工作流程。实验结果表明,应用面向对象的变化检测方法提取地震次生灾害可以取得很好的结果。Remote sensing technology provides valuable data for earthquake hazard management,by which information of seismic secondary disaster such as landslide,avalanche,debris flow,barrier lake can be extracted rapidly and accurately.Applying multi-scale segmentation and context information,the object-oriented fuzzy classification has advantages over traditional methods in accuracy and flexibility.In this paper,we apply this technique to detect changes of Beichuan County,based on pre-and post-event images of Wenchuan earthquake.Four-temporal data was combined with digital elevation model（DEM） to study increment of water body of barrier lake.Also,we explain basic theory of object-oriented classification,summarize the key technique and discuss workflow of seismic secondary disaster extraction.The results show that it is feasible and effective to extract hazard information from image objects.

关 键 词：面向对象分类 地震次生灾害 遥感影像 

分 类 号：P315.9[天文地球—地震学]