论文部分内容阅读
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 SegmentTest)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。
On the basis of studying the traditional geometric correction method, a precise geometric correction method based on automatic point-matching and triangulation with the same name is proposed. The method is based on geometric correction of the image to be corrected through the base map. First, FAST (Features from Accelerated Segment Test) operator is used to extract uniformly distributed candidate feature points on the base map and determine the initial point with the same name by the geo-location information carried by the image itself. After translational error cancellation and cross-correlation bidirectional matching, RANSAC (Random Sample Consensus) coarse subtraction, binary three-point interpolation and other steps to obtain a stable and reliable sub-pixel-level peer-to-peer pairs. Finally, according to the sub-pixel level of the same name point to construct Delaunay triangulation network for image transformation and resampling processing. Landsat satellite ETM is used as baseline datum to carry out geometric precision calibration experiment on environmental satellite CCD data. The accuracy of this algorithm is much higher than the traditional method.