摘要 几何纠正是遥感影像用于防灾减灾救灾、实时动态监测等诸多应用的关键处理环节；同时也是整个遥感图像处理过程中耗时最长、计算最密集的步骤之一。本文在对遥感数字影像的性质和特点、成像模型、坐标系统等做了详细的介绍的基础上，对几何畸变因素、遥感影像几何纠正模型、地面控制点采集方法和几何纠正精度分析做了深入的研究。遥感传感器在接收不同地形类型信息时受到的不同因素的影响不同，地形起伏的变化对遥感影像的精度有很大的影响。遥感影像的几何纠正是遥感影像数据处理的重要环节，所以在影像几何纠正时方法的选用很重要。通过对白银区影像进行相关实验，采用实验对比法，对影像分别使用不同的纠正方法进行整体几何纠正以及单元格网分块、重要点分块几何纠正，并对其实验结果影像进行了分析和总结。实验结果表明进行重要点分块纠正后的影像的精度要高于使用整体纠正方法纠正的精度，而且几何纠正效率高于单元格网分块方法。使遥感影像在实际应用中能够提供更加准确的地理信息，提升其利用价值。本文在研究遥感影像纠正的基础上，提出了按地理特点提取遥感影像重要点构建三角网，将遥感影像分区后再进行纠正的方法。得出将遥感影像图按不同地形、地物区域分块，并根据不同区域的情况选择合适的纠正模型进行几何纠正，分块纠正精度高于整体纠正，实现了遥感影像高精度几何校正。

ABSTRACT Geometrical rectification of the remote sensing images is not only the key process in their applications such as disaster prevention, dynamic detection real-timely and so on, but also one of the longest time-spending and most complex in computing. There is detailed introduction about remote-sensing images’ characteristics, imaging models and geometrical rectification in the paper. What’s more, the factors of geometric distortions, the models of remote-sensing images’ geometrical rectification, the way of collecting control points in the ground, and improvement of geometric accuracy are studied deeply. Different factors’ influence on remote sensors while receiving terrains is different, there is enormous influence on precision of remote sensing images from the topographic relief. As an important part in data processing of remote-sensing images, the geometrical rectification of remote-sensing images is studied in the paper. The Baiyin district are selected to conduct modification test integrally and segmentally. Further, the remote sensing images' partitioning is obtained by comparing the two partitioning ways' precisions. Generation triangulation of the key point is practically modified to get higher-precision and higher-resolution remote-sensing images, which is able to provide us more accurate geographic information and improve the images’ value in use. The geometrical rectification of remote-sensing images is studied in the paper, and the method by that the remote-sensing images are modified after their areal is put forward. Then, particular correction models are confirmed for both the two partitioning methods, and the results show that partitioning rectification's precision is higher than whole rectification's. In conclusion, the remote sensing images are able to be geometrical corrected  accurately in  the  paper.