Vehicle Re-Identification Spanning Ground, Aerial, and Cross-View Domains

Vehicle re-identification (Re-ID) is important for intelligent transportation systems and large-scale surveillance. In this paper, we briefly review vehicle Re-ID methods for ground, aerial, and cross-view scenarios. We organise current approaches into appearance-based models, part-aware methods, transformer-based architectures, and cross-view learning strategies. Our comparison shows that existing methods still struggle with large viewpoint changes and domain shifts between aerial and ground images.