LiDAR-LiDAR Calibration

Overview: Deepen Calibrate is a software tool that makes the critical task of sensor data calibration simple and quick.

1. Calibrations:

  • This page allows the users to create, list, launch and delete calibration datasets. Admins can manage user’s access to these datasets on this page.
  • Click on New Calibration to create a new calibration dataset.

2. Calibration Selection:

  • Upon clicking the New Calibration button, the user can make a selection from the different calibrations. Select the LiDAR-LiDAR Calibration button to create a new LiDAR-LiDAR calibration dataset.

3. Calibration Instructions Page:

  • Upon selecting LiDAR-LiDAR Calibration, the user is welcomed to the instructions page
  • Click on the Get started button after finishing with the instructions. The user has the choice of proceeding with Target or Targetless LiDAR-LiDAR calibration. Choose the appropriate option.
  • If the user choose Target, the user must provide a name for the calibration dataset and dimensions of the target (a rectangular board) based on the sample board in the image. The dimensions have a least count of 1 cm i.e. 0.01 m. Click on Set details to proceed ahead.
  • If the user choose Targetless, the user will only be asked to give a name for the calibration dataset and proceed further by clicking Set details.

4. Calibration Pipeline:

LiDAR-LiDAR Calibration consists of four-five stages based on the type of calibration (target/targetless):
  • Setup details: Add details for your LiDAR sensors.
  • Add files: Upload a single pcd file for each LiDAR sensor.
  • Mapping: Map the uploaded frames by picking exactly four points from each frame.
  • Detect Ground Plane (Only for 'Target'): Get ground planes for better optimisation
  • Calibrate: Run the calibration algorithm of your choice and visualise the results.
A few other details :
  1. 1.
    The user can choose to hit the Save changes option at any stage to prevent loss of work.
  2. 2.
    The user can navigate between the stages and update the details as per convenience.
  3. 3.
    Download the extrinsic parameters by using the Export option.
  4. 4.
    Click on the Help Center button (the top right corner) to get your doubts answered related to the calibration.

4.1 Setup Details:

The user can fill in the names of each LiDAR sensor on this page. Click on Continue to proceed to the ‘Add Files’ stage.

4.2 Add files:

Upload a single file (.pcd format only) for each of the two LiDAR sensors. The calibration is calculated from LiDAR 1 to LiDAR 2. Click on Continue to proceed to the ‘Mapping’ stage.
Note: File name should NOT contain any whitespaces.

4.3 Mapping:

Mapping point clouds helps to determine where a point present in a frame from LiDAR 1 correlates to points in the frame from LiDAR 2. To proceed to calibration, our algorithm requires mapping exactly four points to get accurate results.
The different interactions with the point clouds are:
  • Add marker: enable brush.
  • Erase marker: enable eraser.
  • Clear markers: clear all the points selection.
  • Toolbar: allows rotate, pan, zoom in/out operations.
How to map point clouds using markers?
To start mapping the point clouds:
  1. 1.
    Click the Add Marker button, and click in the area of interest for LiDAR 1.
  2. 2.
    Map the exact points for LiDAR 2 by adding a marker.
  3. 3.
    Repeat the steps for all four markers.
To remove a misplaced marker, click the Erase Marker button and click on the incorrect marker. To remove all the added markers, click on the Clear Markers button.
Mapping for Target Based Calibration:
The four markers selected for the Target based calibration will be the four corners of the target.
The order of the markers should correspond to the length and width configuration in the sample board i.e. length should be in between 1st and 2nd corner while the width should be in between 1st and 3rd corner. Please refer the sample board in the right panel while marking the corner points.
Mapping for Targetless Calibration:
The user may mark any four areas in both the LiDARs which correspond to each other since there is no specific target.
After mapping all the four markers in both the frames, the tool gets quick estimated calibration results. Validate the results in Visualize mode before proceeding to final calibration as these estimates are used to calculate the final extrinsic parameters.
After validating the estimated calibration results, click on Continue to move to the 'Calibrate' stage. If the calibration results seem quite off, then one can use clear markers to clear all markers or Erase marker to erase a specific marker. And retry to correspondence four markers from both lidars.

4.4 Detect Ground Planes (Only for 'Target')

Note: Please go ahead with this step only if your sensors capture ground points else click Continue and move to 'Calibrate' page.
For better results, our system utilises the ground planes to calibrate the sensors in Target based Calibration (This step will not appear in the Targetless calibration flow).
First, Click on Detect ground plane button in the bottom to automatically get the planes in both the LiDARs.
If the user is not satisfied with the plane, they may use Clear ground planes button to remove the detected ground planes and paint the ground points using Add sample button in the top center. The user can also use Erase sample to remove unwanted painted points and use Clear Sample besides Erase sample button to clear all the painted points in both the LiDARs. The user may then click Detect ground plane to get the planes. Click Continue to move on to 'Calibrate' page.
Green points indicate the painted ground samples and Yellow plane indicates the detected ground plane

4.5 Calibrate:

Calibrate with Target:
Click Run calibration button in the bottom to trigger calibration. Upon completion, you will be able to see the extrinsic parameters and the error statistics in the right panel. The extrinsic parameters are from LiDAR 1 to LiDAR 2 while the error stats indicate the rotatio and translation error of the target boards in the two LiDARs.
The user can visualise the fused point clouds by clicking Visualise button on top.
Please note that :
  1. 1.
    Rotation Error indicates the angle between the target planes
  2. 2.
    Translation Error indicates the distance between the center of the boards
Calibrate without Target:
Calibrate your LiDAR sensors by using Generalized Iterative Closest Point(GICP) or Normal Distributions Transform(NDT) algorithm. To find which one to use for your LiDAR sensors, please look in the Help Center (? on top right) section.
Click on GICP or NDT to fetch respective extrinsic parameters. ​​This may take up to 2-3 minutes for us to get more accurate results, depending on the quality of point cloud files.
Validate the results in Visualize mode and mark the calibration process is complete. These extrinsic parameters can be exported by using the ‘Export’ option.
Please note that :
  1. 1.
    The error rate displayed for each algorithm represents the fitness score.
  2. 2.
    We provide a default value for voxel size in the case of the NDT algorithm based on heuristics. The user can wish to choose another value.

5. Extrinsic Calibration Output:

  • roll, pitch, yaw are in degrees and px, py, pz are in meters.
  • roll, pitch, yaw, px, py, pz are the extrinsic parameters downloaded from the calibration tool.
  • lidar1Point is the 3d coordinates of a point in the lidar1 coordinate system.
  • lidar2Point is the 3d coordinates of a point in the lidar2 coordinate system.