卫星线阵立体影像的倾斜仿射变换核线模型

核线约束是构建摄影测量立体模型的重要条件。由于卫星线阵影像的多中心投影特性，难以建立其严格的核线模型。文章从上下视差最小的同名核线特征出发，在仿射变换模型基础上引入核线倾角变化参数，提出一种新的倾斜仿射变换核线模型。通过“无误差”的虚拟同名点仿真试验表明，文章提出的模型受影像范围及地表高差影响较小，在影像区块6 000像元×6 000像元、地表高差1 000m情况下，对于不同形式卫星立体影像的模型误差均小于1个像元。进一步使用实际匹配同名点平差，实验表明：对于SPOT-5、“资源三号”和“高分七号”等卫星影像，文章模型误差相对于匹配观测误差可忽略，实际核线精度在整景影像范围内优于0.8像元，所生成核线影像能够满足后续密集匹配要求。新模型的精确性和普适性相比现有同类模型显著提升，可实际应用于卫星摄影测量处理。

The Oblique-affine Transformation Epipolarity Model for Satellite Linear Array Stereo Imagery

Epipolar geometry constraint is an important condition for constructing photogrammetric-based three-dimensional models. Due to the multi-center projection characteristics of satellite linear pushbroom imagery, it is difficult to find an absolutely strict epipolarity model. In this paper, based on the geometry feature of corresponding epipolar line with the smallest vertical parallax, a new oblique-affine epipolarity model is proposed by introducing the epipolar line inclination parameters into affine transformation model. Through the “error-free”virtual point observation simulation experiment, it is shown that the new model is less affected by the factors such as image block size and the surface height variation. The model errors are all less than 1 pixel for satellite images acquired with different stereoscopic methods. Further, the practical test with real matching corresponding point observations shows that for images such as SPOT-5, ZY-3 and GF-7, the model error is negligible compared to the matching observation error, and the epipolar line precision is able to be better than 0.8 pixels in the whole image range, which guarantee the generated epipolar image meet the requirement of subsequent dense matching. Compared with the existing models, the accuracy and generality of the new model are significantly improved, and it can be practically applied to satellite photogrammetry data processing.