高分辨率SAR图像建筑物地震损毁分析与探测方法研究

Synthetic Aperture Radar (SAR) is one of the most important sensors to monitoring natural disasters, such as earthquakes happening in urban areas.Howerver, in very high resolution(VHR)SAR image, typical buildings are affected by layover, double-bounce, and shadowing effects, which relate to the ranging geometry of radar sensors. It is more compilcated to extract building information from VHR SAR images than medium or low resolution ones. Therefore, new methodologies should be developed to solve this problem. Aiming at that, in this proposal, researches will be carried out as follows: 1) Developing a SAR simulation model to support the image analysis, using three-dimensional models of damaged or destroyed buildings from physics and backscattering simulations. With the simulator, features of the destroyed buildings in VHR SAR image will be analyzed. 2) Based on the feature analysis,detection method with visual perceptual for damaged or collapsed building will be developed. In the procedure, radar imaging geometry will be considered to estimate building image to reduce the false detection. 3) Developing an quantitative model for the building damage ratio estimation based on SAR image.The model will be created by integration of image feature, seismic intensity and other information. In general,the outputs of the proposal will promote applications of the VHR SAR for earthquake damage monitoring in building-up areas.

合成孔径雷达遥感是对城镇地震灾害全天候监测的重要手段，但是在新的高分辨率数据条件下，传统模型和方法已不能解决目标在图像中所表现出的更强内部异质性和不确定性问题。针对这一科学问题，项目将在倒塌损毁/未损毁建筑物的SAR 成像机理研究基础上，开展理论和方法探索。主要包括：1）基于SAR 成像机制，建立倒塌损毁/未损毁建筑物成像模型，分析在不同成像条件下其特征的综合表达及变化规律；2）在特征分析的基础上，建立成像几何约束条件下的视觉感知建筑物震害倒塌损毁探测方法；3）结合多源数据，探索构建基于SAR 影像联合特征的建筑物倒塌损毁定量化描述模型。建立城镇建筑物及震害信息分析与提取的技术体系。通过本项目的研究，可提高损毁区提取的准确性及自动化程度，为地震灾害建筑物损毁监测与评估提供技术及决策支持。

合成孔径雷达（Synthetic Aperture Radar, SAR）是地震灾害全天时、全天候观测的重要手段。项目瞄准高分辨率SAR图像在地震防灾减灾对建筑物倒塌损毁评估应用中，出现的目标在高分辨率图像中所表现出的更强内部异质性和不确定性等问题，通过获取震例遥感数据、结合野外试验调查资料，深入开展研究工作，完成了研究目标与各项研究内容，取得丰富的研究成果。主要包括：1）开展了面向震害信息提取的SAR图像相关预处理研究；2）获取高分辨率SAR图像完好与损毁建筑物的目标对象，结合SAR图像模拟工具，分析了高分辨率数据条件下完好和损毁建筑物的成像特点与图像特征；3）以SAR成像几何机制为基础，提出高分辨率SAR图像完好建筑物的信息提取与高度参数反演方法；4）引入计算机视觉/模式识别方法，结合SAR图像的成像几何特点，提出高分辨率SAR图像倒塌损毁建筑物检测方法；5）结合光学数据、地震烈度等信息，提出联合多特征的SAR图像建筑物损毁程度的定量化判别模型；6）构建了高分辨率SAR图像倒塌损毁建筑物检测技术框架与流程。.总的来说，本项目的研究成果可为地震灾害建筑物损毁监测与评估提供技术及决策支持，提高损毁区提取的准确性及自动化程度。