基于纹理特征的高分辨率合成孔径雷达影像单体建筑物震害信息识别

获取震后建筑物震害信息有利于开展人员救援和灾后重建工作。由于高分辨率合成孔径雷达（SAR）数据少有震前数据存档，利用震后单时相高分辨率SAR数据评估建筑物震害成为研究热点，但利用高分辨率SAR数据对单体建筑物的研究却很少。本文以北川老县城震后0.24mTerraSAR-X聚束模式（ST）数据为数据源，经多视处理后提取建筑物纹理特征，对比分析不同视数大小和纹理计算窗口大小对建筑物震害识别影响，确定最佳纹理计算窗口大小和视数大小。结合震前光学数据，获得SAR单体建筑物轮廓图，随机选取建筑物轮廓样本作为训练样本，引入支持向量机（SVM）和随机森林（RF）分类器识别建筑物震害信息。结果表明，基于纹理特征的SVM、RF方法能有效地识别高分辨SAR影像单体建筑物震害信息，SVM识别精度可达到88.24%，RF识别精度可达到92.47%。可见基于高分辨率SAR数据的纹理特征识别建筑物震害方法稳定有效，可为灾后应急、灾害评估和灾后重建工作提供可靠信息支撑。

Recognition of Single Building Seismic Damage Information in High Resolution Synthetic Aperture Radar Images Based on Texture Features

The building seismic damage information obtained from post-earthquake is useful for rescue and post-disaster reconstruction. Using a single high resolution SAR data to evaluate building damage after earthquake becomes a research hotspot, since high resolution Synthetic Aperture Radar (SAR) data are rarely archived before earthquakes. However, there are a few studies on single building using high resolution SAR data. In this paper, the TerraSAR-X spotlight pattern (ST) with 0.24 m resolution captured post-event of old Beichuan was used as the data source and the texture features of buildings were extracted after SAR multiple look processing. Then the optimal size of multiple look and texture window were derived from comparing and analyzing the impact on building seismic damage classification in different size. The SAR single building footprint was acquired by using pre-earthquake high resolute optical image and selected several samples randomly as training data. Two classifiers, SVM and RF, were utilized to identify building seismic damage information. The result reveals that SVM and RF based on texture features can effectively identify the seismic damage information of single building in high resolution SAR image. The classification accuracy of SVM and RF can reach 88.24% and 92.47%, respectively. It can be seen that the method using high resolution SAR data based on texture feature is stable and effective, and can provide reliable support for post-disaster emergency, disaster assessment and post-disaster reconstruction.