An efficient hyper-graph based tie point matching method for aerial imagery

ABSTRACT For the aerial images characterized by low contrast or containing repetitive patterns or homogeneous textures, robust tie point matching is still a challenging task. In this letter, an effective hypergraph-based tie point matching method is proposed. Firstly, feature points are divided into several point clusters in the overlapping area of images. Secondly, a two-stage matching pipeline, that is, candidate feature point matching and high-order graph matching, is performed. The candidate relationship is utilized to establish the high-order similarity tensor without any information loss. The experimental results show that our algorithm can significantly enhance the matching robustness and success rate compared with traditional tie point matching algorithm, and when compared with state-of-art hypergraph matching methods, our algorithm can obtain approximate equal matching success rate with a largely higher computational efficiency.