严密定位模型辅助的国产卫星影像匹配

Automatic Matching for Optical Imagery Acquired from Domestic Satellites Based on Rigorous Orientation Model

针对不同传感器下的国产卫星影像特点,提出了一种基于严密定位模型的影像匹配方法。在现有方法的基础上,改进扫描行迭代搜索法,实现同名点的快速预测。在全球SRTM数据的辅助下,建立了近似核线方程。通过局部畸变改正,消除了匹配窗口的几何与辐射变形。引入多片最小二乘匹配(MPGC)算法,对匹配结果进行精化,并剔除误匹配点。综合运用了小面元几何纠正法与基于控制网的匹配生长算法,提高了匹配点的匹配精度与分布均匀性。利用天绘一号、资源一号02C、资源三号卫星影像进行试验,结果表明所提出的方法能够较好地结合国产卫星影像特点,实现多传感器下多轨道影像的全自动联合匹配,获得高精度的同名点观测值。

An image matching method based on a rigorous orientation model is presented that exploits the characteristics of optical imagery acquired from domestic satellites from different sensors. Firstly, this paper improves the method for scan line iterative search on the basis of the existing methods, and proposes the rapid correspondence prediction method. Next, with the help of the global SRTM（Shuttle Radar Topography Mission） data, the approximate epipolar line is constructed by projection tracking. The geometric and radiometric deformation in the matching window is eliminated by the correction of local distortion. At last, in original level, the MPGC algorithm is adopted to refine the matching results. This paper proposes a matching algorithm based on the geometric correction of the facet and the matching growth algorithm of a control network, which improves matching accuracy and the distribution of the matching points. A test with imagery data from the TH1,ZY02C and ZY3 satellites showed that the proposed method can combine the characteristics of the optical imagery acquired from domestic satellites, and yield multi-source image matching from multi-source sensors. A well distributed set of matching points can be obtained.