本文選題：高分辨率SAR　切入點：建筑重建　出處：《中國科學技術大學》2016年博士論文　論文類型：學位論文

【摘要】：建筑目標的三維重建是從SAR數(shù)據(jù)／圖像獲得建筑目標的空間三維結(jié)構(gòu)信息的技術,在城市發(fā)展與規(guī)劃、城區(qū)動態(tài)時序監(jiān)測、災后結(jié)構(gòu)與經(jīng)濟損失評估、城區(qū)動目標研究等遙感應用中具有重要意義。本文以米級和亞米級高分辨率SAR數(shù)據(jù)為基礎,以實際環(huán)境中常見的基本立方體及其組合體結(jié)構(gòu)的建筑目標為研究對象,依據(jù)SAR圖像中這些建筑目標的典型電磁散射特征,通過散射特征的準確提取和建筑物幾何參數(shù)的精確估計實現(xiàn)建筑目標主體結(jié)構(gòu)的三維重建。主要開展的研究工作包括：1、建筑目標SAR成像機理及其主要散射特征的研究。系統(tǒng)歸納了建筑目標的三維結(jié)構(gòu)與圖像表現(xiàn)之問的映射關系,討論了其成像特點和影響因素。針對建筑目標SAR圖像中的疊掩(layover)、二次散射(double bounce)、陰影(shadow)等散射特征,詳細分析了它們的形成原因、圖像特點和影響因素。2、建筑目標SAR散射特征提取技術研究。針對傳統(tǒng)提取方法中存在的點擴散函數(shù)高旁瓣干擾問題,從回波域出發(fā),分別提出了結(jié)合Hough變換檢測的二次散射稀疏重構(gòu)技術、結(jié)合TV約束的強后向散射稀疏重構(gòu)技術、結(jié)合幅值反變換和TV聯(lián)合約束的陰影稀疏重構(gòu)技術,降低了點擴散函數(shù)旁瓣對散射特征提取的影響,提高了特征提取精度。仿真結(jié)果和方法對比驗證了三種方法的有效性。3、建筑目標幾何參數(shù)估計技術研究。在SAR數(shù)據(jù)條件較好的情況下,提出了基于立體幾何反演的參數(shù)估計方法,具有良好的估計精度和計算效率。在SAR數(shù)據(jù)條件不理想時,提出了交替迭代與互信息匹配方法,在實現(xiàn)多幾何參數(shù)精確估計的同時,取得了估計性能和計算復雜度的良好平衡。仿真結(jié)果和方法對比驗證了兩種方法的有效性。4、組合體建筑目標的三維重建技術研究。將現(xiàn)有的以基本立方體為主的建筑目標三維重建技術,進一步推廣到更為復雜且更常見的組合體建筑目標,提出了組合體結(jié)構(gòu)填充與匹配優(yōu)化技術。其中,通過提取建筑目標的“結(jié)構(gòu)基元”,利用定位、標記、修正和填充等處理過程構(gòu)建了復雜組合體建筑的結(jié)構(gòu)模型；在此基礎上結(jié)合互信息匹配和多參數(shù)優(yōu)化技術完成了結(jié)構(gòu)匹配,進而實現(xiàn)了高精度的復雜組合體建筑目標的三維重建。在TerraSAR-X數(shù)據(jù)和實際高程測量數(shù)據(jù)的支持下,驗證了所提方法的有效性、可靠性和精確性。5、多視角SAR數(shù)據(jù)下建筑目標的融合重建技術研究。針對單視SAR中建筑目標足跡提取存在的高虛警和高旁瓣干擾問題,利用多視角SAR數(shù)據(jù)提出了基于“稀疏重構(gòu)—融合—置信評分檢驗”的足跡提取方法。針對復雜場景中建筑目標存在后向散射混疊、結(jié)構(gòu)遮擋以及結(jié)構(gòu)未知等問題,提出了基于結(jié)構(gòu)模型的構(gòu)建與融合的方法。仿真分析和實測數(shù)據(jù)驗證了上述兩種方法在應對各自問題中的有效性和可靠性。
[Abstract]:Three-dimensional reconstruction of building targets is a technique to obtain spatial three-dimensional structural information from SAR data / images in urban development and planning, urban dynamic time series monitoring, post-disaster structure and economic loss assessment. In this paper, based on the high resolution SAR data of meter level and sub-meter level, the research object of basic cube and its combination structure in practical environment is taken as the research object, which is of great significance in the application of remote sensing, such as the research of moving target in urban area. Based on the typical electromagnetic scattering characteristics of these building targets in SAR images, Through accurate extraction of scattering features and accurate estimation of geometric parameters of buildings, 3D reconstruction of the main structure of building targets is realized. The main research work includes: 1, SAR imaging mechanism of building targets and its main scattering characteristics. Research. The mapping relationship between 3D structure and image representation of building target is summarized systematically. In this paper, the imaging characteristics and influencing factors are discussed, and the formation reasons of them are analyzed in detail according to the scattering characteristics of the SAR images of building targets, such as overlapped layover, double double bouncees, shadow shadow, and so on. Image characteristics and influencing factors. 2. Research on SAR scattering feature extraction technology of building target. Aiming at the problem of high sidelobe interference of point diffusion function in traditional extraction methods, the echo domain is used to solve the problem of high sidelobe interference. The second scattering sparse reconstruction combined with Hough transform, the strong backscattering sparse reconstruction with TV constraint, the shadow sparse reconstruction with amplitude inverse transform and TV joint constraint are proposed respectively. The influence of the sidelobe of the point diffusion function on the extraction of scattering features is reduced. The accuracy of feature extraction is improved. The effectiveness of the three methods is verified by the comparison of simulation results and methods. The geometric parameter estimation technology of building target is studied. When the SAR data condition is good, A parameter estimation method based on solid geometry inversion is proposed, which has good estimation accuracy and computational efficiency. When the SAR data condition is not ideal, an alternative iteration and mutual information matching method is proposed, which realizes the accurate estimation of multiple geometric parameters at the same time. A good balance between estimation performance and computational complexity is achieved. The effectiveness of the two methods is verified by comparing the simulation results with the methods. 4. The 3D reconstruction technology of the composite building target is studied. 3D reconstruction technology for building targets, Further extended to the more complex and more common composite building targets, an optimization technique for structural filling and matching of composite structures is proposed, in which, by extracting the "structural elements" of the building targets, positioning and marking are used. The structural model of complex composite building is constructed by modifying and filling processes, and the structure matching is completed by combining mutual information matching and multi-parameter optimization techniques. Furthermore, the 3D reconstruction of complex composite building object with high precision is realized. The validity of the proposed method is verified by the support of TerraSAR-X data and the actual height measurement data. Reliability and accuracy. 5. Research on fusion reconstruction technology of building target based on multi-view SAR data. Aiming at the problems of high false alarm and high sidelobe interference in single view SAR, high false alarm and high sidelobe interference are involved in the extraction of building target footprint. Based on the multi-view SAR data, a footprint extraction method based on "sparse reconstruction, fusion and confidence score test" is proposed. In view of the problems of backward scattering aliasing, structural occlusion and unknown structure in complex scenes, this paper proposes a new method of footprint extraction based on "sparse reconstruction, fusion and confidence score test". The method of building and fusion based on structure model is proposed, and the effectiveness and reliability of these two methods in dealing with their respective problems are verified by simulation analysis and measured data.
【學位授予單位】：中國科學技術大學
【學位級別】：博士
【學位授予年份】：2016
【分類號】：TN957.52
，

本文編號：1561239