基于相关变化检测与面向对象分类技术的 多源遥感图像震害信息提取

遥感图像面向对象分类作为空间信息提取的关键技术， 在震害信息提取方面发挥着非常重要的作用， 然而由于光学遥感影像是正射图像， 只能提取建筑物屋顶信息， 这使得单一利用震后光学影像进行震害信息提取存在一定的局限性. 针对该问题， 本文提出了一种基于合成孔径雷达(SAR)相关变化检测的光学影像震害建筑物面向对象提取方法， 即在光学影像面向对象提取的数据中融合SAR相关性， 对光学影像进行面向对象提取震害建筑物时不仅考虑建筑物的几何、 光谱等特征， 还加入震前震后变化信息即SAR相关性进行分类. 在此基础上， 选取2008年汶川M_S8.0地震震区都江堰地区作为研究区进行试验. 结果表明， 本文提出的方法相对于单一使用光学影像进行震害建筑物提取, 其准确度有较明显的提高. 

Extraction of earthquake damage buildings from multi-source remote sensing data based on correlation change detection and object-oriented classification techniques

Object-oriented classification technology of remote sensing images, as a spatial information extraction method, plays a key role in earthquake damage information extraction. However, it has limitations in extracting buildings from optical remote sensing images due to the characteristic vulnerable to weather and other reasons. To solve this problem, this paper proposes a method for detecting building damage by optical image in combiniation with SAR correlation changes and object-oriented classification, which extracts buildings from fusion data including optical image and SAR correlation image. Not only the spatial and spectral features, but also the correlation of buildings is considered during the extraction. The M_S8.0 Wenchuan earthquake caused a wide range of building’s collapse and casualties. Taking Dujiangyan area near the source as an example, the method above is tested. The results show that the accuracy of building extraction is improved by using the proposed method compared with the method only from optical image.