Sensor Calibration

The Aerial Calibration preset in LiDARSnap to make global roll, pitch, and yaw adjustments to a dataset by modifying the sensor-to-IMU transforms in the lidar Settings->Calibration tab:

When building a point cloud, the lidar's orientation is determined by taking the IMU's orientation and applying a set of transforms (rotations and translations) specific to the lidar scanner. A lidar dataset will exhibit consistent and systematic errors if these transforms are incorrect:

Typically sensor-to-IMU translations are taken from mechanical drawings, and do not need to be solved for using LiDARSnap. Generalized sensor-to-IMU rotations can also be determined from mechanical drawings (e.g., 180 degrees along IMU-X, -180 degrees along IMU-Z), however precise values, unique to each system, must be determined via a calibration routine.

All lidar systems are calibrated prior to delivery to the end-user, however frequent use, mishandling, and system age can over time invalidate this initial factory calibration, so relative accuracy in a dataset may be improved via sensor calibration.

Sensor calibration requires certain geometry to be present in a dataset. Man-made planar features need to be present, ideally with a variety normals. IMU-to-sensor pitch and roll can be solved for using flat ground (such as a parking lot), however solving IMU-to-sensor yaw requires upright and pitched surfaces.

When running the Aerial Calibration LiDARSnap preset, the Sensor tab will indicate that the Mounting Rotation (IMU-to-lidar yaw, pitch, and roll) has been enabled for optimization, as well as certain laser intrinsic values:

The Trajectory tab of LiDARSnap should have nothing enabled when using the Aerial Calibration preset. In general, it's not recommended to have both sensor and trajectory features enabled for optimization in a single LiDARSnap run - either solve for trajectory parameters OR sensor parameters.

The result of an aerial calibration LiDARSnap run is modified sensor-to-IMU rotation values (as well as modified laser intrinsic values), visible in the lidar settings calibration window. Often, lidar data misalignment issues persist even after sensor calibration, as non-constant errors in the trajectory are not resolvable via sensor calibration.

Last updated