Topology-Aware Handover for Multi-UAV Vehicle Tracking
A new framework addresses trajectory fragmentation in multi-UAV vehicle tracking for intelligent transportation systems. The system, described in arXiv preprint 2605.15779, introduces a lightweight Topology-Based Spatiotemporal Handover mechanism to maintain global identity persistence across multiple drone fields of view. It overcomes limitations of appearance-based re-identification in nadir views by using spatial topology and temporal cues, enabling real-time multi-camera multi-vehicle tracking. The approach is designed to support network-level analysis such as origin-destination estimation, which is hindered by current isolated data silos. A high-throughput parallel pipeline ensures scalability for continuous monitoring.
Key facts
- arXiv:2605.15779v1
- Announce Type: cross
- Abstract presents a real-time Multi-Camera Multi-Vehicle Tracking (MCMT) system
- Introduces lightweight Topology-Based Spatiotemporal Handover mechanism
- Addresses visual ambiguity and computational cost of appearance-based Re-ID in nadir views
- Implements high-throughput parallel pipeline
- Designed for global identity persistence across multi-UAV Fields of View (FOV)
- Aims to enable network-level analysis like Origin-Destination estimation
Entities
Institutions
- arXiv