After calibration, it is important to assess its quality. One way is to apply some 3D reconstruction, mapping or SLAM method and see the results. However, for a quantitative measure we propose a validation procedure that can be easily applied by using a mobile robot equipped with a 2D laser range finder.
Although this step is optional, it is important to actually measure the effect of the calibration procedure.
Each dataset has the following structure:
ROOT:
calibrated
corner.data
central.data
undistorted
corner.data
central.data
raw
corner.data
central.data
RAW: uncalibrated sensor
UNDISTORTED: data acquired with the old procedure (IAS-13 paper)
CALIBRATED: data acquired with the new procedure that takes into account the biasing phenomena
Each file is composed by two columns, like:
1065.84 1081.63
1065.84 1081.47
1065.84 1079.26
1065.84 1081.36
1065.84 1080.01
1065.84 1083.95
1065.84 1079.53
1065.84 1079.36
1065.84 1079.2
... ...
The first column are the laser measurements (ground truth), the second column are the estimated distance perceived by the sensor