面向震后极化SAR图像震害识别的震区多类型建筑物提取研究

Earthquakes are one of the most destructive natural disasters. In all kinds of hazards caused by the earthquake, the building damage is the major cause of casualties. After an earthquake, rapidly, accurately and overall obtaining building damage information of the earthquake-stricken areas can help to effectively guide the implementation of the emergency rescue and can reduce disaster losses and casualties. In order to quickly and accurately identify the building earthquake damage information, This study is mainly on the damaged building and undamaged building extraction from a single post-earthquake full polarimetric SAR (PolSAR) image. Previous studies generally lack the consideration of part-collapsed buildings. Therefore, the buildings of earthquake-stricken region are going to be divided into intact parallel buildings, intact oriented buildings, parallel residual walls, oriented residual walls and completely collapsed buildings in this study. Some characteristics of the five types of buildings will be analyzed in this study, such as geometric construction, spatial distribution and polarization scattering characteristic, and so on. According to the analysis results, the sensitive feature set is going to be established. Two kinds of research methods will be used to extract the buildings of earthquake-stricken region in this study, which are the method based on fusion of polarimetric and texture features in PolSAR image and the method based on construction of different building scattering models. According to the extraction results of the five kinds of buildings, the building damage index will be calculated for the building earthquake damage assessment. The different PolSAR data sources in different areas are going to be selected to validate the applicability of the proposed method. This study can extend the application of damage assessment for PolSAR data, and produce remarkable practical significance and academic value.

地震是最具破坏力的自然灾害之一。建筑物损毁是震后人员伤亡的主要因素。快速、准确、全面地获取建筑物震害信息能够有效指导应急救援的实施，减少人员伤亡与灾害损失。考虑数据获取方便且数据信息量丰富，本项目仅使用震后一景极化SAR数据识别建筑物震害信息。鉴于以往研究普遍缺少对部分倒塌建筑物的考虑，本项目拟将震区建筑物分为完好平行建筑、完好方位建筑、平行残留墙体、方位残留墙体和完全倒塌建筑五种类型，对这些建筑的几何结构、空间分布、极化散射特性等进行剖析，建立敏感特征集。拟采用两种研究思路实现多类型震区建筑物的提取，即融合多种极化特征与纹理特征和构建震区不同建筑物的散射模型。基于震区多种建筑类型的提取结果，依据震害指数对灾区建筑物进行震害评估。本项目拟选择不同震例的不同极化SAR数据源对提出方法的适用性进行验证。本项目能够拓展极化SAR在震害评估中的应用，具有显著的现实意义和学术价值。

地震是最具破坏力的自然灾害之一。建筑物损毁是震后人员伤亡的主要因素。及时有效的建筑物震害信息能够指导应急救援的高效实施，减少人员伤亡与灾害损失。为了快速准确识别建筑物震害，本项目使用震后一景极化SAR数据对倒塌建筑与未倒塌建筑的识别方法与技术进行了多方位深入研究。在极化SAR图像中，除了极化信息，纹理特征也是非常重要的信息，而震区不同类型建筑物具有比较明显的纹理特征差异，为此本项目将变差函数作为纹理特征算子，提出了结合变差纹理的建筑物震害识别方法，取得了84.54%的倒塌建筑识别精度和80.18%的未倒塌建筑识别精度。多种纹理特征的识别可靠性比一种纹理特征高，为此本项目提出了基于精度加权的多特征融合方法，融合了四种纹理特征，同时综合应用最优极化对比度增强方法与极化分解模型提取建筑物震害信息，获得了81%的总体评估精度，相比传统方法精度提高了9%。为了降低倒塌建筑漏识带来的震害评估风险，本项目将震区建筑物细化为完好平行建筑、完好倾斜建筑、倒塌平行墙体、倒塌倾斜墙体和完全倒塌建筑五种类型，提出两种新的极化特征参数λ_H和H_λ，利用这两种参数逐层提取震区建筑，相比以往研究这种方法从未倒塌建筑中再次分离出2种类型的倒塌建筑，获得了83.3%的建筑物震害信息提取精度，有效提高了建筑物震害识别的准确性。SAR图像中经常存在大量的高亮山脊线，弱化了建筑区的辨识度，为了降低山脊线对建筑区快速识别的影响，本项目利用Chan-Vese模型提取山脊线轮廓，结合极化分解方法提出了基于变分原理的山脊线识别方法，取得了84.09%的山脊线识别精度，明显优于传统方法。本项目提出的多项建筑物震害识别新技术与新方法，显著提高了建筑物震害识别精度，拓展了极化SAR数据在地震领域的应用。本项目还构建了包含27种特征的倒塌建筑SAR图像特征库系统，能够为相关研究人员提供丰富的研究样本，提高科研人员的研究效率，还能为震害评估人员提供指导意见。因此，项目研究成果具有显著的科学意义和社会价值。

内容获取失败，请点击重试