Reproduce “GL-DT: Multi-UAV Detection and Tracking with Global-Local Integration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Juanqin Liu, Leonardo Plotegher, Eloy Roura, Shaoming He
Researchers from the Beijing Institute of Technology and the Autonomous Robotics Research Centre (TII) developed GL-DT, a real-time system for multi-UAV detection and tracking that integrates a global-local detection strategy with a robust tracking algorithm. The system achieved the lowest ID Switch (33) and highest MOTA (82.84) on the MOT-FLY dataset, alongside robust performance on a new challenging FT dataset with ultra-small targets.
Key insights
- Integrating spatio-temporal features, rather than relying solely on appearance, is crucial for accurate and stable detection of fast-moving, small UAVs against dynamic backgrounds.
- An adaptive global-local detection strategy balances broad scene understanding with fine-grained localization, significantly improving detection accuracy and computational efficiency for small targets.
- Robust data association combining geometric, motion, and relational costs with a probabilistic memory recovery mechanism effectively minimizes identity switches and maintains trajectory continuity, even during occlusions or lost detections.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.