节点文献

临近空间大倾角遥感图像几何校正方法研究

Researches on Geometric Correction of Near Space Large Inclined Angle Remote Sensing Images

【作者】 徐庆阳

【导师】 张晔;

【作者基本信息】 哈尔滨工业大学 , 信息与通信工程, 2009, 硕士

【摘要】 随着卫星技术和计算机技术的飞速发展,数字遥感卫星影像已经在各行各业中发挥着越来越大的作用。从遥感影像中提取地球空间信息,需要把遥感影像投影在某一固定的参照系统中并修正原始影像所存在的几何变形,以便进行影像信息的几何量测、相互比较和复合分析。如何将遥感影像精确地投影到规定的参照系统中、准确消除原始影像所存在的几何变形是遥感影像处理和应用的一项关键技术。临近空间,是对海拔20千米到100千米空间范围的一个通用性称谓。长期以来,各国军方大都忽视了临近空间的军事价值,近年来却纷纷开始重视起来,开展了对临近空间的相关技术研究,临近空间大倾角遥感图像几何校正就是其中之一。本文针对临近空间大倾角遥感图像几何校正技术中的难点问题进行了深入研究,提出了分段多项式校正模型,并通过迭代去除错误控制点算法和均匀分布算法优化了SIFT兴趣算子自动选取控制点的过程,实现了全自动几何校正。本文首先研究了产生遥感图像几何失真的原因,总结出六方面的因素:传感器成像方式,传感器外方位元素,地形起伏,地球曲率,大气折光和地球自转,并得出了在临近空间条件下图像分辨率随倾角的变化规律。接下来本文对几何校正模型进行了研究,首先比较了仿射变换校正模型和一般多项式模型,介绍了几何校正的原理和步骤,接着分析了3种重采样算法的优点和不足。最后针对临近空间大倾角失真遥感图像提出了一种分段多项式几何校正方法,取得了理想的校正精度。在几何校正过程中,控制点的自动选取和准确匹配是影响校正精度的关键。本文在最后研究了几何校正过程中控制点选取的方法,并重点研究了控制点自动选取算法。首先分析了控制点数量对校正模型精度的影响,接着比较了Forstner兴趣算子和SIFT兴趣算子,并研究了错误控制点的迭代去除算法和均匀分布算法。通过迭代去除错误控制点对和均匀分布算法,控制点得到了优化,校正精度得到了显著的提高,并将其利用于分段多项式几何校正算法,达到了亚像元的校正精度。

【Abstract】 With the great development of satellites and computer technology, digital remote sensing image is taking a more and more important place in many fields. For geometrical measuring, reciprocal comparison, image compound analysis and etc., the remote sensing images always need to be expressed at a certain projection reference system and eliminate geometric distortion of imagery. How to project remote sensing images onto a certain reference system and keep geometrical position, shape, dimension and orientation of objects unchanged, namely geometrical rectification of remote sensing images, is an essential process in remote sensing image’s processing and application. Near space is the space scope from altitude 20km to 100km. For a long time, most countries’military ignored the military value of near space. However, they have begun to pay attention to it, and make scientific researches on technology in near space, the research on geometric correction of near space oblique remote sensing images is one of them.In this dissertation, an in-depth study of the theory of near space remote sensing image geometric correction is performed. For geometric correction of oblique remote sensing images, the piecewise polynomial correction model is studied, then use the technique of iterative error control points removal and the technique of uniform distribution to optimize the process of control points selection using SIFT operator, and achieve the automatic geometric correction .This dissertation first studies how the remote sensing picture geometry is distorted, and the dissertation summarizes that there are six aspects. They are sensor imagery mode, sensor element of exterior orientation, hypsography, curvature of the earth, atmosphere refraction and earth rotation, and obtains the image resolution with the changes of angle in the near space conditions.Then the dissertation studies the geometric correction models. First, compares the affine transformation model and the general polynomial correction model, introduce the principle and steps of the geometric correction method, and analyses the strengths and weaknesses of the three kinds of interpolation methods. Finally, for geometric correction of near space oblique remote sensing images, a piecewise polynomial geometric correction method is studied, and achieved the desired accuracy of calibration.Detecting control points automatically and accurate matching are the two crucial steps in geometric correction. Finally, the dissertation studies how to choose control points, especially the automatic geometric correction. How the number of control points impact the accuracy of geometrical rectification is studied, then compares Forstner operator and SIFT operator. After that, use the technique of iterative error control points removal and the technique of uniform distribution to optimize the process of control points selection, the accuracy of the correction has been improved, and achieve the sub-pixel accuracy of the correction using these techniques and the piecewise polynomial geometric correction method.

  • 【分类号】TP751
  • 【被引频次】2
  • 【下载频次】314
节点文献中: 

本文链接的文献网络图示:

本文的引文网络