基于优化视觉词典的震后高分遥感影像震害建筑物检测

在缺少震前参考信息前提下,提出了一种基于优化视觉词典的震后高分遥感影像震害建筑物检测方法。 首先通过WJSEG(wavelet-JSEG)分割以及一组非建筑物筛选规则提取潜在建筑物集合;其次利用光谱、纹理及几何形态学特征构建了一种震害视觉词典模型,跨越了从像素到震害特征间的“语义鸿沟”;在此基础上设计了一种基于类内和类间惩罚因子的视觉词典优化策略,减少了信息冗余及证据冲突;最后通过随机森林分类器将建筑物进一步划分为完好建筑物、部分震害建筑物及废墟。 在两组实验中,该方法的总体精度均达到 85%以上,从而可为震后应急响应救援及灾后重建提供关键的决策支持信息。

Damaged building detection based on optimized visual dictionary from post-earthquake high-resolution remote sensing images

Being lack of the pre-earthquake reference information, a new method of damaged building detection of high-resolution remotesensing image based on optimized visual dictionary is proposed. Firstly, wavelet-JSEG (WJSEG) segmentation and a set of non-buildingscreening rules are applied to extract the potential building set. Secondly, a visual dictionary model of earthquake damage is constructedby introducing spectral, texture and geometric morphological features to across the semantic gap between pixels and earthquake damagefeatures. On this basis, a visual dictionary optimization strategy based on intra-class and inter-class penalty indexes is designed to furtherreduce redundant information and evidence conflict. Finally, the buildings are further classified into intact buildings, partially damagedbuildings and ruins by random forest classifier. In two experiments, the overall accuracy of the proposed method reached more than85%, which can provide key decision support information for post-earthquake emergency response and reconstruction.