利用自动匹配与三角剖分进行遥感图像几何精校正

在研究传统几何精校正方法的基础上，提出了一种高精度的基于自动同名点匹配和三角剖分技术的几何精校正方法，该方法是通过基准底图对待校正图像进行几何精校正的。首先利用FAST (Features from Accelerated Segment Test)算子在基准底图上快速提取均匀分布的候选特征点，通过图像自身携带的地理定位信息确定初始同名点对；经平移误差消除、互相关双向匹配、RANSAC(Random Sample Consensus)粗差剔除、二元三点插值等步骤获取稳定可靠的亚像元级同名点对；最后根据亚像元级同名点对构建Delaunay三角网进行图像变换和重采样处理。以Landsat卫星ETM为基准底图对环境卫星CCD数据进行几何精校正试验，本算法几何精校正精度较传统的方法得到了很大提高。

Remote sensing image geometric accurate rectification based on automatic matching and triangulation

This paper proposes a Geometric Accurate Rectifi cation method in remote sensing image based on reference image library. Firstly, uniformly distributed feature points are obtained in reference image using FAST (Features from Accelerated Segment Test) algorithm, the original matches of feature points are found based on the geographical information. Secondly, the highprecision matches of points are searched out in the input image through translation error removing, exact matching and RANSAC (Random Sample Consensus) gross error elimination. Finally, Delaunay triangulation is constructed for image transformation and resampling. It is verifi ed by HJ Satellite experimental images, compared with traditional method, the accuracy of proposed method has been greatly improved.