LiDAR-to-Vehicle Calibration of Arbitrary Sensor Setups via Object Reconstruction

To begin, prepare the PCD files (in the sensor frame) for each sensor, following the example provided in the /input directory.
The corresponding vehicle poses must be specified for each pcd in the /input/ego_poses.yaml file.
For the poses, we expect the coordinate system (x-forward, y-left, z-up).
The poses follow the schema
- x (translation)
- y (translation)
- z (translation)
- x (quat)
- y (quat)
- z (quat)
- w (quat)

The config file currently has four entries:

• Initial Transformations: Represents the initial roll, pitch, yaw angles as well as the translation vector of both sensors.
• Target position: The parameters min_bound and max_bound define the global position of the target in all point clouds.

A Dockerfile is provided for CaLiV:

You can build the Docker file with a defined <tag>:

docker build -t caliv:latest -f docker/Dockerfile .

Then run the calibration with:

docker run -v $(pwd)/output:/app/output -it  caliv:latest

The results, namely the corrected LiDAR transformations are saved in the /output path.