A real-time fiducial tag system for LiDAR point clouds, robust to lighting and compatible with visual markers like AprilTags.
LiDARTag is a real-time fiducial tag system designed for LiDAR point clouds. It provides a robust alternative to image-based markers by functioning reliably in any lighting condition, including complete darkness. The system detects tags, estimates their pose with high accuracy, and decodes unique IDs, enabling applications like object tracking, sensor calibration, and SLAM.
Robotics researchers and engineers working with LiDAR sensors, especially those involved in autonomous systems, SLAM, multi-sensor calibration, and perception in challenging lighting environments.
Developers choose LiDARTag because it solves the critical limitation of lighting sensitivity in fiducial marker systems. Its real-time performance, high accuracy, and compatibility with existing visual markers make it a unique tool for robust multi-sensor perception where cameras alone are insufficient.
This is a package for LiDARTag, described in paper: LiDARTag: A Real-Time Fiducial Tag System for Point Clouds
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Functions reliably in complete darkness, dingy conditions, and rapidly changing light, overcoming a key weakness of image-based markers like AprilTags, as demonstrated in the provided GIFs.
Processes data at 100 Hz, faster than typical LiDAR sensor frequencies, enabling real-time applications such as SLAM and dynamic object tracking.
Achieves millimeter-level translation error and a few degrees rotation error, validated by motion capture systems, making it suitable for precise calibration tasks.
Uses a reproducing kernel Hilbert space (RKHS) method to decode tag IDs with 99.7% accuracy, minimizing false positives in cluttered environments.
Designed to work alongside visual fiducial markers like AprilTags, facilitating sensor fusion and extrinsic calibration, as highlighted in the linked calibration paper.
Requires specific dependencies like ROS Melodic, TBB, and NLopt, with manual compilation steps and potential Eigen version issues, making initial deployment time-consuming.
Custom marker sizes necessitate regenerating function dictionaries via MATLAB scripts, adding overhead for adaptation beyond the default sizes (0.61m, 0.805m, 1.22m).
The README admits that a more detailed introduction is pending, and users must rely on GitHub issues for support, indicating incomplete or evolving documentation.
Tightly coupled with ROS and specific LiDAR sensors (e.g., Velodyne), limiting portability to non-ROS systems or newer hardware without significant modification.