LAMM: Large-scale invariant multi-source remote sensing image matching based on Adaptive Many-to-many Mapping

With the rapid development of integrated space–air–ground Earth observation technologies, multi-source image matching has become a crucial prerequisite for satellite–UAV data fusion, large-scale mapping, and cross-platform environmental monitoring. However, large-scale differences between image pairs often lead to sparse correspondences, while viewpoint variations further intensify geometric inconsistencies under large-scale scenarios, making it difficult for existing methods to obtain sufficient and high-quality correspondences. To address these challenges, this paper proposes a large-scale-invariant multi-source image matching framework based on an adaptive many-to-many mapping. Specifically, we first introduce a many-to-many matching scheme that alleviates correspondence sparsity under significant scale differences by relaxing geometric constraints. Then, we construct a multi-association difference filtering module that effectively removes mismatches caused by insufficient constraints while preserving reliable many-to-many relations. Finally, we develop an adaptive bidirectional mapping mechanism that dynamically adjusts the search strategy during fine matching according to local scale variations, thereby eliminating reliance on prior scale assumptions. Extensive experiments conducted on various multi-source datasets demonstrate that the proposed method achieves superior robustness and accuracy compared to state-of-the-art approaches. Our code is provided at the following repository: https://github.com/yiqing18/LAMM