一种用于制作CBERS-02C卫星融合影像的配准方法

A registration method used for fusion of CBERS-02C satellite images

提出了一种用于制作资源一号02C卫星(CBERS-02C)融合影像的配准方法.首先对多光谱及全色影像进行几何精纠正,并采用SIFT算法对两种纠正影像的缩略图进行粗匹配;然后通过提取Harris特征点,并根据粗匹配的结果预测对应同名点的位置,使用基于灰度相关的匹配方法进行精匹配;待剔除匹配点粗差后,再采用基于三角网的面元纠正实现影像的精确配准.多组数据的试验验证了所提方法的可行性.该方法已成功应用于CBERS-02C卫星数据的自动化处理系统.

This paper proposes a registration method for generating fused images from CBERS-02C images.The process can be divided into three steps.Firstly,the panchromatic and multispectral images are precision-rectified,and then the thumbnails of the precision-rectified images are matched coarsely by using SIFT-based method.Secondly,Harris-featured points are extracted,and the coordinates of their corresponding points are calculated according to the results of coarse matching,then the precise matching is performed based on normalized cross correlation.Thirdly,after the elimination of gross errors,the image rectification is conducted based on the Triangulated Irregular Network （TIN） to achieve high-precise registration.The feasibility of the method is tested by using several groups of data to do the same experiment.This method has been successfully applied to the automatic data-processing system of CBERS-02C satellite.