局部特征匹配的卫星影像几何偏差估计可行性分析

几何定位偏差是评估卫星影像质量的重要参数之一。本文利用特征匹配方法进行卫星影像几何偏差估计，并对该方法的可行性进行分析。首先，选取包含典型地物的局部卫星影像构建路标基准影像库；其次，采用局部点特征描述子对卫星影像与路标基准影像库进行亚像素提取与匹配，确定同名像点，然后为提升同名像点匹配精度，提出了顾及空间关系与几何一致性约束的匹配策略，继而，对卫星影像与路标基准影像库的匹配像点进行内插计算，获取同名像点的地理空间坐标；最后，通过差值计算得到卫星影像相对路标基准影像的偏移参数，完成几何偏差估计和可行性分析。本文以我国风云卫星为例，选取包含海岸线、湖泊、山脉、河流及岛屿的局部卫星影像集构建路标基准影像库，并选用SIFT、SURF、ORB 3个局部特征进行测试分析。测试结果表明，基于局部特征匹配策略对卫星影像几何偏差估计的系统误差小于0.1像素，定量地证明了局部特征匹配方法对卫星影像几何偏差估计的可行性。

Feasibility analysis of using local feature matching for the geo-reference offset estimation of satellite images

Geo-reference offset is one of the important parameters for evaluating the quality of satellite images. In this paper, we estimate the geo-reference offset of satellite images using local feature matching, followed by the analysis of its feasibility. Firstly, the subset of satellite images covering typical ground landmarks are selected as the reference database. Following that, the local point features are extracted for searching for the correspondences between satellite image and the reference database at sub-pixel scale. In order to improve the matching accuracy, the constraints of spatial relation and geometric consistency are taken into the consideration. Next, the interpolation of the correspondences between the satellite image and the reference database is performed to obtain the spatial coordinates of the tie-points in the satellite images. Finally, the rotation and offset parameters of the satellite image relative to the reference database are obtained with quantitative analysis of its feasibility. In this paper, we choose Fengyun satellite for the experiments, where the image subsets including coastline, lakes, mountains, rivers and islands landmarks are selected as the reference database. Three local features, in terms of SIFT, SURF and ORB are used for local feature matching. The experimental results show that the RMSE error of estimating the geo-reference offsets is less than 0.1 pixels, which confirms the feasibility of the local feature matching strategy.