Improving Image Matching Between Digital Twin and Real-World Using Cross-Domain Geo-Localization Techniques

Digital twins, as virtual replicas of real-world entities, have gained attention for their ability to synchronize with reality and facilitate simulations. However, inherent differences between digital twin and real-world datasets challenge traditional image-based matching methods. This paper evaluates the performance of cross-domain image geo-localization methods applied to digital twins and the real world, focusing on visual and semantic feature matching. Experimental results show that both methods perform well, with semantic feature matching outperforming visual feature matching, indicating its advantage in handling differing visual characteristics between domains. These findings confirm the potential of semantic feature matching for effective cross-domain matching between digital twins and the real world.