【作者】 赵福军；

【导师】 张景发；

【作者基本信息】 中国地震局工程力学研究所 ， 防灾减灾工程及防护工程， 2010， 博士

【摘要】 近年来,随着遥感影像分辨率的提高和遥感信息提取技术的发展,遥感技术逐渐成为快速获取地震灾情信息、震后应急和震害快速评估的有效手段。但以往的遥感震害信息提取结果精度较低、震害识别对象单一、没有形成比较完备的遥感地震应急软件平台,基于以上问题,本文研究遥感影像震害信息提取的方法和实际应用。本文的主要研究成果如下:1.研究了基于面向对象分类和震害知识库的遥感影像震害信息提取方法。分析了遥感影像面向对象信息提取技术所涉及的多尺度分割、影像对象特征的定量描述和模糊分类技术。在影像分割的尺度参数选择上,提出了基于影像对象内部标准差和局部空间自相关Geary系数的最优分割尺度选择算法,该算法可确定一幅影像上不同地物的各自最优分割尺度,试验数据的分类精度评价结果验证了该算法的准确性和有效性。建立了遥感震害知识库、阐述了遥感震害知识库的组织结构和主要内容。提出了基于遥感影像面向对象变化检测的震害信息提取技术流程,为实现不同典型震害的信息提取和定量计算提供了方法上的依据。2.提出了不同破坏等级建筑物震害信息提取的影像特征参数优选方案和推理规则。全面分析了遥感影像上建筑物的震害特征。明确了建筑物遥感震害等级与宏观地面调查的建筑物震害等级的对应关系,指出了两者判别依据的不同。通过理论分析和试验结果研究,提出了不同破坏等级建筑物震害信息提取的影像特征参数优选方案和推理规则,不仅为建筑物遥感震害知识库的建立提供了依据,而且提高了建筑物震害信息提取精度。并以海地、伊朗巴姆地震为例,实现了利用面向对象分类和面向对象变化检测方法提取不同震害级别的建筑物信息。3.提出了矢量与影像叠加的面向对象变化检测道路震害信息提取方法。研究了遥感影像上道路、桥梁震害级别与宏观地面调查震害级别的对应关系、判别依据和差异性。分析了地震前后道路和桥梁的影像特征,提出了道路和桥梁震害识别的特征参数优选方案和推理规则。在道路震害信息提取上,提出了矢量与影像叠加的面向对象变化检测道路提取算法,该算法实现了在缺少震前遥感影像的不利情况下,利用道路矢量数据和震后遥感影像定量地提取出道路震害信息。研究了高分辨率全色卫星影像和航空影像上桥梁震害信息提取方法。4.给出了典型次生灾害识别的特征参数优选方案和上下文特征判别依据,实现了面向对象变化检测的典型次生灾害信息提取。详细总结并分析了滑坡、崩塌、泥石流、堰塞湖等典型地震次生灾害在多种遥感影像上的影像特征,得出地物的空间关系特征是识别次生灾害的重要依据,并以此给出了典型次生灾害识别的推理规则。阐述了震后影像上典型次生灾害的识别过程和利用震前震后遥感影像的面向对象变化检测方法提取典型次生灾害的具体实现过程。5.用雷达影像干涉相干变化指数法判别地震破坏程度时,提出了地震破坏等级划分的阈值选取方法。研究了雷达影像的相关性分析法和干涉相干变化指数法。通过对汶川地震的20个地震灾区的试验研究,得出相关性分析法是一种判别地震破坏程度的宏观评价方法。在用雷达影像干涉相干变化指数法判别地震破坏程度时,提出了地震破坏等级划分的阈值选取方法,克服了以往阈值选取的主观性,利用该方法得出的地震破坏等级图,经实际地面调查验证发现,其结果更快速有效。6.提出了遥感地震应急与震害快速评估技术流程,分析了该技术流程的多个关键技术及其软件实现。建立了集分布式遥感数据库系统、遥感震害信息提取与震害快速评估系统、遥感影像网络发布系统于一体的遥感地震应急综合处理平台。利用该软件平台快速获得了汶川地震后的遥感震害信息提取结果,并以此绘制了基于遥感数据的汶川地震烈度分布图,该图与基于地面调查结果的汶川地震烈度分布图基本一致,从而验证了遥感地震烈度判定方法的有效性。更多还原

【Abstract】 With the emergence of high resolution remotely sensed imagery and development of remote sensing information extraction technology in recent years, remote sensing, RS for short, became an effective means for acquiring seismic disaster information rapidly, emergency response and seismic disaster assessment. Due to lower accuracy of seismic disaster information extraction, singularity of seismic disaster recognition target and lack of complete remote sensing seismic emergency response software platform, this dissertation study on seismic information extraction method using remotely sensed imagery and its practical application. The main works of this dissertation are summarized as follows:1. Seismic information extraction method based on seismic disaster knowledge base and object-oriented image classification techniques are studied thoroughly.The key techniques of object-oriented information extraction are analyzed, including multi-resolution segmentation, quantitative description of image object features and fuzzy classification. The optimal segmentation scale model is presented using image object standard deviation and local Geary’s C, which could make sure different scale factor for each ground object and is exact and efficient after validation. Furthermore, knowledge base about remote sensing seismic disaster is established. The structure and content of knowledge base are also described. Seismic information extraction work flow based on object-oriented change detection technique are presented, which would provide a basis for extracting seismic disaster information.2. Feature optimal scheme and reasoning rules are pointed out to extract building seismic damage information from different damage level buildings.Building seismic characteristics are thoroughly analyzed on remote sensing image. The corresponding relation between RS and macro-ground investigation building seismic damage level is confirmed, at the same time it is pointed that they are different from each other in judgment basis. Through theoretic analysis and tests, feature optimal scheme and reasoning rules are pointed out to extract building seismic damage information from different damage level buildings. Thereby, RS seismic damage knowledge base is established initially according to that scheme and rules, and thus the building seismic damage extraction accuracy is improved. In this dissertation, taking Haiti and Bam earthquake as example, different damage level building information are obtained using object-oriented classification and object-oriented change detection method.3. A new road extraction algorithm based on object-oriented change detection method using road vector data overlaid on post-earthquake image is presented.Difference judgment basis are studied as well as the corresponding relations between RS and macro-investigation damage level about road and bridge. Feature optimal scheme and reasoning rules are presented by analyzing the characteristics of road and bridge on RS images. A new road extraction algorithm based on object-oriented change detection method using road vector data overlaid on post-earthquake image is presented. This algorithm can realize quantitative analysis on road seismic damage when it is lack of pre-earthquake images. High-resolution pan image and airborne image are employed to study bridge seismic information extraction method.4. Feature optimal scheme and context feature of typical seismic secondary disasters are presented, typical seismic secondary disasters information extraction is accomplished using object-oriented change detection methodThe characteristics of multi-source remote sensing image of typical seismic secondary disasters, which include landslide, collapse, debris flow and landslide-dammed lake, are summarized. It is pointed that ground objects’spatial relationship features is the important judgment basis to recognize secondary seismic disasters. Consequently, reasoning rule was brought forward to recognize seismic secondary disasters. The procedures of seismic secondary disasters information extraction are illustrated using post-earthquake image only and pre- and post-earthquake images with object-oriented change detection method.5. A new threshold selection method for different seismic damage level is presented when using radar images’interferometric coherence change index.The methods of correlation analysis and interferometric coherence change index using radar image are analyzed. It is proved that correlation analysis method is a macro-assessment method for seismic disaster by experimental study on the selected 20 example areas in Wenchuan earthquake. A new threshold selection method for different seismic damage level is presented when using radar images’interferometric coherence change index. This method can overcome subjectivity of threshold selection. Seismic damage level map is draw using this method and validated by ground investigation. Moreover, it is proved that this threshold selection method is more rapid and efficient than ground investigation.6. The technique procedure of RS earthquake emergency response and rapid assessment was presented,several key technologies related to this procedure and its software implementation are analyzed.Distributed database system, RS seismic disaster information extraction and rapid seismic damage assessment system and web publishing system of remotely sensed imagery are integrated in the remote sensing earthquake emergency response software platform. Based on that platform, RS seismic disaster information is obtained rapidly about Wenchuan earthquake. The seismic intensity distribution map based on remotely sensed imagery is draw and validated that it is accordance to Wenchuan seismic intensity distribution map based on ground investigation. Consequently, the effectiveness of RS seismic intensity determination method is validated.更多还原