Event
S3KF: A Spherical State-Space Panoramic 3D Multi-Object Tracking Framework
1 reports · 1 sources
Zhongyuan Liu et al. propose S3KF, a panoramic 3D multi-object tracking framework based on a motor-rotating LiDAR and four fisheye cameras. It models target bearing on the unit sphere S² using a two-degree-of-freedom tangent-plane parameterization, and jointly estimates it dynamically with box scale and depth. The authors report that in experiments on self-collected real-world sequences, the framework achieves decimeter-level planar tracking accuracy, outperforms 2D panoramic baselines in identity continuity in dynamic scenes, and runs in real time on the Jetson AGX Orin platform.
Report timeline
- S3KF: Spherical State-Space Kalman Filtering for Panoramic 3D Multi-Object Tracking
The authors propose S3KF, a panoramic 3D multi-object tracking framework based on a motorized rotating LiDAR and a quad-fisheye camera rig. It models object bearing on the unit sphere S^2 using a two-degree-of-freedom tangent-plane parameterization and jointly estimates it with box scale and depth dynamics.
arXiv cs.RO ·