【作者】 王伟；

【导师】 童小华；

【作者基本信息】 同济大学 ， 地图制图学与地理信息工程， 2008， 硕士

【摘要】 随着遥感影像数量和种类的不断增加,不同遥感影像间信息的集成利用研究体现出越来越重要的作用。基于不同种类高分辨率遥感影像的立体定位及地物提取是遥感信息集成应用的重要部分。本文以QuickBird遥感影像数据和框幅式航空遥感影像数据为基础,研究了高分辨率推扫式航天影像和框幅式航空影像的联合立体定位、地物提取以及二者间的核线关系。研究的主要工作归纳如下:(1)介绍了常见的遥感传感器严密传模型和通用模型,以及基于传感器模型的立体定位原理。重点研究了QuickBird遥感影像立体定位的系统误差的规律以及不同纠正模型的纠正精度。实验证明,仿射纠正模型的纠正效果最好。在此基础上,研究了基于QuickBird遥感影像和航空遥感影像的联合立体定位,通过实验比较发现,联合定位的精度高于QuickBird遥感影像的定位精度,低于航空遥感影像的定位精度。(2)阐述论证了投影轨迹法核线模型和共面方程法核线模型的关系。在共面方程思想的基础上,推导出了推扫式遥感影像的核线方程、以及推扫式遥感影像与框幅式遥感影像间的核线方程,并通过实验研究了QuickBird遥感影像的核线以及QuickBird遥感影像和航空遥感影像间的核线的形状、性质以及精度。实验证明,推扫式遥感影像上的核线是核曲线,且在较小范围内近似于直线,不同点对应的核曲线相交,但并不交于同一点。基于共面方程的核线模型具有亚像素级精度。(3)研究了基于QuickBird遥感影像与框幅式遥感影像的建筑物的提取方法,主要包括基于SUSAN算子的边缘检测、基于Hough变换的边缘提取及基于不变矩和核线约束的匹配,并在此基础上提出了建筑物提取的策略方法。论文针对传统SUSAN算子的缺点做出了改进,提出了基于双阈值SUSAN算子的边缘检测,实验证明,基于双阈值SUSAN算子的边缘检测算法在一定提取信息量的基础上有效提高边缘检测的精确性。本文还通过实验验证了不变矩的不变性,并结合基于共面方程核线模型核线约束,提高了匹配的速度和可靠性。论文最后提出了建筑物提取的策略和步骤,并通过实验实现了基于QuickBird遥感影像与框幅式遥感影像间建筑物的提取。最后,本文对所做工作进行了总结,并对未来需要继续研究的工作做了展望。本文得到了国家自然科学基金项目(40771174)、国家863计划课题(2007AA12Z178)、“曙光计划”项目(07SG24)的资助。更多还原

【Abstract】 With the increase of quantity and types of remote sensing imagery, integration of different types of remote sensing imagery has become more and more important. Geopositioning and object extraction are important parts of research on integration of different types of remote sensing imagery. This paper does research on geopositioning, epipolarity and object extraction based on integration of the linear CCD push-broom imagery and frame perspective imagery with QuickBird images and aerial images as experimental data. The main works are summarized as below:1. This paper does research on the systematic error, accuracy assessment of error refinement based on different refinement models and geopositioning based on integration of the linear CCD push-broom imagery and frame perspective imagery. It is proved in the experiments that affine model has the highest accuracy and geopositioning accuracy based on integration of QuickBird image and aerial image is less than geopositioning accuracy based on QuickBird images and higher than geopositioning accuracy based on aerial images.2. This paper does research on shape, property and accuracy of epipolarity of the linear CCD push-broom imagery and epipolarity between the linear CCD push-broom imagery and frame perspective imagery. It is proved in the experiments that the shape of epipolarity in linear CCD push-broom image is curve. Different epipolarity intersect at different points and the accuracy of epipolarity is above one pixel.3. This paper does research on methods for extracting buildings, which mainly includes edge detection based on SUSAN operator, line extraction based on Hough Transform and matching method based on invariant moment and epipolarity constraint. Extraction strategy based on QuickBird images and aerial images is presented in the end. The deficiency of SUSAN operator is partly improved by double-threshold SUSAN operator which is proposed in this paper. It is proved in the experiments that the accuracy of edge detection is improved while information content detected is not reduced. Invariability of invariant moment is also proved and epipolarity constraint is applied for improving speed and reliability of matching in the experiment. Extraction strategy based on QuickBird images and aerial images is proposed in the paper and applied in the experiment.In the end of the paper, the work which has been done in the paper is summarized, and the problems requiring further research are discussed.更多还原