Repository including Gazebo models, plugins and worlds to test algorithms for extrinsic calibration of lidar-camera pairs. Package developed at Intelligent Systems Laboratory, Universidad Carlos III de Madrid.
- Bumblebee XB3 Camera
- Velodyne VLP-16
- Velodyne HDL-32
- Velodyne HDL-64
This repository includes several sensors and calibration target models to evaluate the performance of extrinsic calibration of lidar-camera pair in the Gazebo Simulator.
Note: The models included in this repository were designed for evaluating the LIDAR-camera calibration algorithm described in [1], whose code is provided here.
Sensors:
- Bumblebee XB3 Camera (Left - center only)
- Velodyne VLP-16 (Based on DataspeedInc)
- Velodyne HDL-32 (Based on DataspeedInc)
- Velodyne HDL-64 (Since 3D meshes are not available, those of HDL-32 model are used instead)
Calibration targets:
- Calibration pattern with wood (maple) texture
- Calibration pattern with chessboard texture
- Chessboard planes for recreating worlds required to test the KIT Calibration Toolbox
- Velodyne plugin providing PointCloud2 with same structure as driver (x, y, z, intensity, ring) and simulated Gaussian noise. (Code from DataspeedInc, although minor patch for vertical resolution issue is included)
- Gazebo can take up to 30 seconds to load the VLP-16 pluggin, 60 seconds for the HDL-32E, and much more HDL-64E
- Gazebo cannot maintain 10Hz with large pointclouds
- Solution: User can reduce number of points in urdf
roslaunch velo2cam_gazebo real_stereoVLP16_trans.launch
[1] Guindel, C., Beltrán, J., Martín, D. and García, F. (2017). Automatic Extrinsic Calibration for Lidar-Stereo Vehicle Sensor Setups. IEEE International Conference on Intelligent Transportation Systems (ITSC), 674–679.
Pre-print available here.