A multi-view image matching algorithm for feature point underthe moving Z-Plane constraint
移动高程平面约束的多视影像特征点匹配

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