Reproduce “Real-time 6-DoF Pose Estimation by an Event-based Camera using Active LED Markers” 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
Gerald Ebmer, Adam Loch, Minh Nhat Vu, Germain Haessig, Roberto Mecca, Markus Vincze, Christian Hartl-Nesic, Andreas Kugi
A system for real-time 6-DoF pose estimation leverages an event-based camera and active LED markers, achieving a mean latency of 354 µs and output rates of 3.81 kHz on a desktop PC. It delivers translational errors of 34.5 mm (±16 mm) and orientation errors of 0.74 degrees (±0.15 degrees) at distances up to 4.8 m, outperforming ORB-SLAM2 in indoor drone experiments and tracking markers during outdoor flights at 10 m/s.
Key insights
- Strategic tuning of event camera biases significantly enhances the signal-to-noise ratio and enables distinct event front detection for ALMs, simplifying the event processing pipeline.
- Leveraging a priori knowledge of ALM blinking frequencies allows for robust and rapid identification of individual LEDs within the sparse event stream, even under fast motion.
- An asynchronous tracking and pose estimation pipeline, decoupled from initial detection, is critical for achieving extremely low latency and high output rates, surpassing the performance of traditional frame-based systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.