基于传感器几何特性和图像特征的影像配准方法

“天绘一号”卫星是我国第一代传输型立体测绘卫星，搭载有三线阵CCD相机、多光谱相机和2m分辨率全色相机。多光谱相机波段间影像的配准是保证多光谱影像质量和有效应用的前提，在现有配准方法的基础上，文章提出了基于传感器几何特性和图像特征的配准方法来改善影像配准效果。这种方法首先基于相机几何特性对待配准图像进行调整，然后对图像进行局部多点取图，采用尺度不变特征变换（Scale Invariant Feature Transform，SIFT）法提取图像特征并进行初步匹配，利用匹配点领域灰度相关性来进行精匹配和误匹配点的剔除，最后获得图像位移差，进而完成图像配准。实验结果表明，这种方法能够提高多光谱波段间影像的配准成功率，准确率达到95％以上，大幅改善了匹配精度。

A Improved Matching Method for Image Based on Sensor Geometric Features and Image Features

TH-1 Satellite is the first generation of transmission-type three-dimensional mapping satellite, equipped with a three-line array CCD camera, multi-spectral camera and 2-meter resolution panchromatic camera.To improve matching accuracy between different bands of TH-1 Satellite＇s multispectral image, this paper introduces an improved method based on geometric features of multi-spectral sensor and Scale Invariant Feature Transform （SIFT） features of image. First, image is adjusted according to geometric features of multi-spectral sensor. Second, different parts of image are extracted and SIFT feature points picked up. The pixels gray of feature point＇s space are used to calculate similarity of gray value to process accurate match. Then we get the distance of relative positions of images and finish the matching process. The experiment shows that this method can improve the correct match rate between different bands of multi-spectral image to more than 95%, and the matching accuracy is also much advanced.