基于SAR多特征变化检测的震害建筑物提取研究

发布时间：2018-03-20 03:30

本文选题：SAR　切入点：面向对象　出处：《中国地震局工程力学研究所》2017年博士论文　论文类型：学位论文

【摘要】：强地震引起严重伤亡和财产损失。地震作为一种全灾种的自然灾害,既会导致房屋倒塌、道路中断等直接损失,也会引起一系列次生灾害。相对于光学遥感来说,合成孔径雷达(Synthetic Aperture Radar,SAR)应用于地震灾害的监测与评估具有独特的优势:受雨雪等天气状况影响小、侧视成像能够更为全面地反映建筑物信息。然而由于传感器、地域的不同,提取方法往往不具有较强的普适性。遥感影像的变化检测技术可以利用震前震后的多时相影像提取变化信息,是解决这个问题的主要方法之一。本文以SAR影像变化检测技术中的几个关键问题为研究目标,开展基于多特征的SAR变化检测震害建筑物提取研究,取得的主要成果如下:(1)梳理分析了不同震害类型、不同走向建筑物在不同入射角、不同分辨率和不同极化方式SAR影像的成像机理,重点对建筑物的散射特征、极化特征进行了分析与总结。进而分析了SAR影像的特征和震前震后的变化特征,在此基础上进行了变化检测特征优选。(2)引入面向对象的方法,以影像分割对象为研究目标,采用单特征和多特征融合的方法,实现了高精度的建筑物震害信息提取,克服了基于像素的SAR影像相关性和相干性变化检测方法提取震害建筑物精度不高的问题。(3)发展了基于极化SAR影像的雷达遥感震害指数,提出了双极化雷达遥感震害指数DualRRSDI,进而对震前震后双极化SAR影像进行变化检测;基于极化协方差矩阵服从复Wishart的分布的理论,构建了全极化SAR影像似然比模型,选择完好建筑物与震害建筑物样本,对极化似然比图像进行分类。(4)采用面向对象的思路,综合强度相关性、相位相干性与极化似然比等SAR影像多特征的变化信息,开展建筑物震害信息提取研究。(5)总结出基于多特征的SAR变化检测震害建筑物提取技术流程,根据数据的不同极化方式、是否具备干涉能力等不同情况进行不同特征的组合,进一步明确了地震应急阶段利用SAR影像变化检测进行震害建筑物提取的工作模式。
[Abstract]:Strong earthquakes cause serious casualties and property losses. As a kind of all-disaster natural disasters, earthquakes not only lead to direct losses such as collapse of houses and interruption of roads, but also cause a series of secondary disasters. Synthetic Aperture radar (SAR) has a unique advantage in earthquake disaster monitoring and assessment: it is less affected by weather conditions such as rain and snow, and side-view imaging can more comprehensively reflect building information. The change detection technology of remote sensing image can extract the change information from the multitemporal image before earthquake. It is one of the main methods to solve this problem. This paper aims at several key problems in SAR image change detection technology, and carries out the research on extracting earthquake damage building based on multi-feature SAR change detection. The main results obtained are as follows: (1) analyzing the imaging mechanism of SAR images with different incidence angles, different resolution and different polarization modes, and focusing on the scattering characteristics of buildings, according to the different types of earthquake damage, different strike buildings, different incident angles, different resolution and different polarization modes. The polarization features are analyzed and summarized. Then, the characteristics of SAR images and the changes after earthquakes are analyzed. On the basis of this, the change detection features are selected, and the object-oriented method is introduced to segment the objects of the images. The method of single feature and multi-feature fusion is used to extract the seismic damage information of buildings with high precision. This paper overcomes the problem that the accuracy of extracting seismic damage buildings is not high in pixel based SAR image correlation and coherence detection method. It develops the radar remote sensing damage index based on polarimetric SAR image. The dual-polarization radar remote sensing damage index DualRSDI is proposed to detect the changes of dual-polarization SAR images before and after the earthquake, and based on the distribution theory of polarization covariance matrix from complex Wishart, the likelihood ratio model of fully polarized SAR images is constructed. Selecting samples of intact buildings and earthquake damaged buildings, classifying the polarimetric likelihood ratio images. 4) using object-oriented thinking, synthesizing the variation information of multi-features of SAR images, such as intensity correlation, phase coherence and polarization likelihood ratio, etc. Carrying out the research of building damage information extraction. (5) summing up the SAR damage detection technology flow based on multi-feature, combining different features according to different polarization mode of data, whether has interference ability or not, etc. The working mode of seismic damage building extraction using SAR image change detection in earthquake emergency stage is further clarified.
【学位授予单位】：中国地震局工程力学研究所
【学位级别】：博士
【学位授予年份】：2017
【分类号】：P315.9;TN957.52

【参考文献】