Typical disaster damage target extraction method based on object oriented classification

摘要: The extraction of typical targets (such as water, road and damaged buildings) nature disaster is important to the emergency management, post-disaster damage assessment monitoring. This paper shows an objected oriented method extract building in the earthquake Wenchuan(2008), Yushu(2010) and Haiti(2010). We built optimal feature sets with spectral features, shape texture features context features. Experiment result that this can flood area, buildings effectively achieve a relatively high accuracy. These experimental studies are leading opportunity integrate classical survey and image oriented semi-automatic interpretation.