【作者】 李关龙；

【导师】 张晔；

【作者基本信息】 哈尔滨工业大学 ， 信息与通信工程， 2010， 硕士

【摘要】 遥感摄影测量能够获取大量的、大范围的资料数据,并且在采集信息时受时空等条件限制较少,现在已经成为地球环境信息提取的重要手段。随着空间遥感技术的发展,在获得极其丰富的信息的同时,越来越强烈要求对多角度、多传感器甚至是多平台之间获取的信息数据进行信息互补,来弥补单一遥感影像信息的不足,基于多角度遥感图像对目标进行三维空间信息提取就是其中的一种形式。随着三维信息提取技术的发展,人们对物体三维空间信息描述的应用要求越来越高。这也促进了三维图形技术和科学计算可视化技术的快速发展,让更多的研究人员把目光投向了三维可视化的研究上。基于获得真实纹理细节信息的目标及场景的三维可视化技术,由于视觉效果好、真实感强,进而得到了人们的亲睐并在众多领域应用广泛。本文主要研究三维信息的获取以及这些数据模型的具有真实纹理信息的可视化应用。首先,本文研究了多角度遥感像的立体匹配方法,采用Harris算法与SIFT特征匹配算法相结合的方法提取匹配立体像对中的同名点,提出了基于遥感成像理论的同名点对匹配方法。利用传感器参数模型对多角度遥感图像进行目标及场景的三维信息提取,在此过程中利用泰勒级数将传感器参数成像模型线性化,并通过误差迭代的方式对参数进行精细标定,从而提高了对目标及场景三维信息提取的精度,增强了系统的鲁棒性。然后,本文利用基于逐点插入法的Delaunay三角剖分和规则网格剖分相结合的网格剖分方法,再通过多项式插值和多项式拟合针对建筑物和桥梁等典型目标以及地形进行几何建模。完成几种典型目标模型构建的同时建立了几种几何模型基元模块。最后,本文对纹理信息提取和映射方法做了研究,提出了基于成像模型的纹理匹配映射方法,完成目标模型与纹理信息的匹配映射,增加了三维目标模型的纹理信息。采用透视变换和消隐处理等计算机图形学原理将纹理信息数据与三维空间数据合并一起进行可视化显示和多角度漫游观察,实现多角度观察和分析的目的,增强了目标及场景的真实感观。本文中做了大量的半实物及实物仿真实验并进行了实验验证。更多还原

【Abstract】 Remote sensing photogrammetry has a lot of advantages, such as large range and large amount of information accessing, less restricted conditions. Therefore, it becomes to be an important way of information extraction on the resources and the environment. With the continuous development of space remote sensing technology, we can obtain much more information. At the same time, people increasingly strong demand for accessing to the complementary of information which is from multi-angle, multi-sensor or even multi-platform to fill the shortage of single remote sensing image. The three-dimensional information extraction based on multi-angle remote sensing images is an example.With the development of the three-dimensional information extraction technology, people require more and more applications on three-dimensional spatial information of the objects. It promotes the development of three-dimensional graphics technology and the visualization technology in scientific computing. And a growing number of researchers are paying attention to the visualization of three-dimensional information. In this process, we try to get texture information as much as possible in order to enhance realistic view of objectives and scenarios. As is widely used, three-dimensional visualization becomes to be one of the hot issues.This paper mainly studies on three-dimensional information accessing and the visualization of these data with real texture information. First, this article mainly researches stereo matching method in multi-angle remote sensing images by Harris ang SIFT. And accomplishes the three-dimensional information extraction based on remote sensing imaging theory. In this process, the paper resolves the problem about fine calibration of sensor parameters by Taylor series linearization and error Iteration. It improves the accuracy of extraction and enhances the stability of the system.Then, it uses Delaunay triangulation, polynomial interpolation and polynomial fitting to get the geometric model of typical targets and terrain, such as buildings bridges and scenarios. Additionally, there are basic element models to build complex models. At the same time the models are proposed texture mapping method that based on remote sensing imaging model. It increases the texture information of objectives and scenarios.Finally, this article solves the visual display of data and roaming display of the three-dimensional reconstruction by using computer graphics principles. The final texture information data combined with the three-dimensional data for visualization and multi-angle roaming. So that the realistic view of three-dimensional target model is enhanced. Thereby, it carries out multi-angle observation and analysis purposes, and in this paper a lot of semi-physical or physical simulation experiments are done to verify them.更多还原