移动高程平面约束的多视影像特征点匹配

A multi-view image matching algorithm for feature point underthe moving Z-Plane constraint

针对影像匹配中由于重复纹理和遮挡导致的不确定性问题,提出移动高程平面约束的多视影像特征点匹配方法。该方法将多视影像上提取的特征点投影到不同高程的物方格网平面,利用平面上格网单元来约束候选同名特征点,对多视影像进行选择性匹配,避免部分影像遮挡对匹配的影响。实验选取UCX(UltraCamX)数字航空影像进行匹配,取得可靠的匹配结果,从而验证了本文方法的有效性。

Considering the uncertainty in image matching caused by repetitive textures and occlusions, this paper presents a multi-view image matching algorithm for feature point under the moving Z-Plane constraint. This algorithm projects the feature points extracted from multiple images onto different height planes, then uses the grid cells in the Z-Plane to constrain the cor- responding matching candidate. It carries on the selective matching for multiple images depending on the projection rays in the grid cell, so as to avoid the problem caused by occlusions. Finally, the validity of the algorithm proposed in this paper is verified by the experiments using four UltraCamX （UCX） digital aerial images, the algorithm is shown to have reliable matching results.