【作者】 王伟；

【导师】 李霖；

【作者基本信息】 武汉大学 ， 地图制图学与地理信息工程， 2017， 博士

【摘要】 高精度立体测绘卫星影像的分析与计算是测绘遥感信息科学的重要研究领域,是遥感应用的重要基础。随着遥感应用的深入和精细化,高精度、多时相、多视遥感影像的获取和处理成为迫切的需求。其中,针对多视、多时相遥感影像的应用无疑是急需解决的关键技术之一。论文立意于我国高分辨率立体测绘卫星发射初期,本文的研究为初期试验和后续论证服务提供了有力的支持,多年的应用也验证论文方法的有效性和可用性。论文在充分分析高分辨率卫星遥感影像成像建模与几何辐射处理及应用技术国内外研究现状基础上,针对我国发射高分辨率立体测绘卫星的重大需求,重点研究了高分辨率立体卫星影像处理和提取地物要素的诸多关键技术,包括:1)高精度立体测绘卫星影像对地定位理论与技术方法,论述了高分辨率卫星遥感成像模型构建,重点分析了立体卫星遥感严密成像模型和有理函数模型建立的技术方法与模型特点,并利用资源三号立体测绘卫星影像开展了严密成像模型和有理函数模型对地定位处理与验证试验,全面地剖析了大区域卫星控制延拓特征规律,为卫星遥感影像高精度几何处理奠定了基础理论与技术方法;2)研究了高精度卫星影像辐射质量提升理论与技术方法,论述了高分辨率卫星成像辐射传输模型构建的基本理论方法,重点分析了多视卫星成像相对辐射校正、交叉辐射校正、MTFC复原等关键技术,利用资源三号立体测绘卫星影像开展了多种因素影响下的传递函数分离与补偿技术试验,验证了系统传递函数补偿与分离补偿的适用性与有效性,有效提高了高分辨率立体测绘卫星的影像辐射质量；3)结合测绘应用中对地物边界的敏感性,利用鲁棒性保持的超分辨率重建技术,有效地提高地物边界的辨识度,将资源三号卫星的分辨率从3.6m提高到0.52m,信息量提升10%,为影像数据分析提供有力支撑;4)开展针对高分辨率立体卫星影像提取建成区的实验,提出基于多核学习的多特征融合的建成区检测方法,针对高分辨率立体测绘卫星影像,实施协同检测建成区的方法,并在我国不同地区和不同特征的区域开展实验,验证高分辨率立体卫星影像在城市土地变化和城市发展变化中的应用。论文详细论述了面向立体卫星遥感影像地物信息提取的精确几何定位与深度辐射信息提升处理的关键技术,实现了立体卫星遥感成像高精度几何成像模型与辐射传输模型的构建;实现了基于多视、多时相卫星遥感影像的多种地表信息及其变化信息的精确提取,充分挖掘了立体卫星数据的应用效能。本文创新点主要包括:1)提出了基于立体遥感卫星影像交叉校正的影像辐射质量综合提升方法;2)实现了基于有理函数模型与区域网平差技术的立体遥感影像立体模型定向解算与地形信息提取,融合鲁棒性保持的超分重建技术,高效提高地物辨识度,并开展了技术方法试验验证;3)针对立体测绘卫星影像,充分利用前视、正视和后视影像,发展多种纹理提取方法,提出基于多核学习的多特征融合的建成区检测方法,发展针对高分辨率立体测绘卫星影像的协同检测城市建成区方法,并开展实验验证。本文研究对立体卫星遥感数据的精确解析与处理、信息的深入挖掘与应用具有重要价值。更多还原

【Abstract】 Ground object extractions and their changes based on remoting image are the important research fields.Recently,with the development of remote sensing technology,high resolution and multi angle stereo remote sensing,as well as,ground object information extraction becomes key techniques.How to process,analyze and apply stereo remote sensing images is a new and difficult science issue.Multi angle stereo remote sensing images based ground object change extraction could be applied in different social and economic life areas,such as mapping and geographic information updating,city managing and planning,agriculture investigating and so on.In this thesis,after the state of arts of high resolution remote sensing images’modeling and processing,focuses are put on key technologies:1)high precision stereo satellite remote sensing ground positioning methods.The paper discusses high resolution remote sensing imaging model,describe the stereo satellite images’ rigid imaging model and rational function model and its ground positioning processing,and explain the control data’s reducing rule in large range block adjustment.2)High precision radiometric quality enhancing theory and technology.The paper discusses satellite remote sensing’s radiometric transfer function’s basic theory,investigate multi angle stereo remote sensing images’ atmosphere correction,relative radiometric correction,absolute radiometric correction and MTF compensation,by using ZY3 stereo remote sensing images MTF’s separate and compensation experiments have been carried out,which proved stereo remote sensing images MTF compensation’s important effect.3)satellite remote sensing image based ground objects extraction.The paper implements several key techniques,including stereo remote sensing images’orientation,DSM extraction,ground slope and aspect extraction and ground object change detection and so on.4)Built-up area extraction based remote images.The paper provides a multi-kernel learning method to fuse multi-features to extract built-up areas,and this cooperative detection can enhance the efficiency of built-up area extraction.The innovations in the thesis include the following aspections.The paper investigates high accurate geometric and radiometric processing method and algorithm based on stereo satellite remote sensing images,realizes stereo images’geometric and radiometric modeling,and studies several ground object information extraction method and experiments,and develop stereo remote sensing images’proficiency.This study is significant useful for stereo remote sensing images’ accurate processing and information extraction.更多还原