Vehicle Lidar (3D and 2D) Calibration
PreviousData Collection for Vehicle-2D Lidar calibrationNextData Collection for Vehicle Lidar Targetless Calibration
Last updated
Last updated
Calibration Process:
1. Calibrate List Page:
Click on ‘New Calibration’ on the calibration list page.
2. Calibration type selection:
Click on Vehicle-LiDAR calibration, to start it.
3. Calibration welcome page:
A welcome page with the set of instructions is shown at the beginning, just click on ‘Get started’ to go forward.
4. Vehicle configuration page:
Enter the calibration configuration along with the calibration name. Following are the vehicle configurations.
Wheelbase: Distance between left/right front wheel center and left/right rear wheel center in meters. If asked for a Wheelbase without left/right, measure either the left or right wheelbase. If the wheelbase is the same on both left and right, the right wheelbase is optional.
Track: Distance between front/rear left wheel center and front/rear right wheel center in meters. If asked for just Track without front/rear, measure either the front or rear track.
Diameter: Distance between bottom of the front/rear wheel to the top of the front/rear wheel (wheel height) in meters. Note: wheel diameter must include tires as well.
Overhang: Distance between center of the front/rear wheel to the front/rear bumper of the vehicle in meters.
5. Setup Lidar Details:
Enter the name of the lidar and the distance of the board from the vehicle. Also enter the lidar height if measured. This is an optional field. Steps will change depending on whether the value has been given or not.
For 360 degree lidar, all 4 side the boards can be placed.
For directional lidars, the field of view of the lidar can be selected. For the front view lidar, rear board is not required. Similarly for others.
Negative values can be configured in the board distance . For example if the front board is placed parallel to the front side of the vehicle. It can be placed ahead of the front side or behind the front side. If it is placed in the front side then the distance between the board and the front side of vehicle is entered in positive values. If the board is placed behind the front side of the vehicle then the negative values are to be entered.
Distance from LiDAR to ground: This configuration is required only when there are no ground points visible in the lidar. In such cases, the accurately measured distance from the lidar to the ground along the line perpendicular to the ground can be entered to proceed with the calibration process. Note: Do not set this value if ground points are visible in the point cloud. Setting this value can reduce calibration accuracy.
6. File Upload for front-view:
User needs to upload a pcd file which has a calibration target on the front side of the vehicle.
7. Detecting the ground plane and board plane for front-view:
When lidar-height is provided: When lidar-height is provided, only the board plane needs to be detected. So just mark some board points and click on detect plane to detect the board plane.
When lidar-height is not provided:When lidar-height is not provided, both the ground and the board plane are to be detected. If the ground points are more, then most probably the ground points get auto-detected or else mark the ground points to be detected by the ground plane. Similarly, board points should be marked to detect the board plane.
After ground plane detection
After board plane detection
Use the Add Sample/ Erase Sample option to add/erase the ground points, which are used in ground plane detection. Painted points are denoted in the green color.
Use the Add Marker/ Remove Marker option to add/erase the ground points, which are used in board plane detection.Painted points are denoted in the red color.
8. File Upload for left-view:
Similarly upload a file for left-view.
9. Detecting ground plane and board plane for left-view:
Similarly detect ground plane and the board plane for left-view.
10. File Upload for right-view:
Similarly, upload a file for right-view.
11. Detecting ground plane and board plane for right-view:
Similarly, detect the ground plane and the board plane for right-view.
12. Run calibration:
This step has the run calibration button. On clicking, calibration runs. This usually gets finished in 1 or 2 mins.
The extrinsic parameters can be viewed to the right.
Visualization:
Visualization shows the planes which are detected and also the lidar and the vehicle coordinates. This also shows the vehicle corners on the ground. The planes can be deselected. Also the fused point cloud can be deselected, to just look at the vehicle and the lidar axes and the vehicle body.
Error statistics:
Plane distance error is the mean of the distance between the plane LiDAR points to the respective planes.
There are planes detected based on the configuration like front board plane, left board plane, right board plane, rear board plane and ground plane.
Each plane has associated 3D points in the point cloud which are either painted or auto-detected.
After running calibration, the distance between the 3D points to the corresponding plane is calculated and the mean of all the distances is shown in the error statistics.
Export:
Extrinsic parameters can be exported to a json file using the button at the top right corner.
After clicking "Get Started" on the Vehicle Lidar Calibration welcome page, select "2D LiDAR" option to calibrate 2D lidar sensor.
Provide the details of the vehicle such as the vehicle type, tracks, wheelbase and overhangs. (Similar to vehicle lidar calibration for 3D sensor as described above)
Along with vehicle configuration, user should give the dimension of the target's edge. The target is an equilateral triangular pyramid (Tetrahedron).
Provide details about the lidar's direction with respect to the vehicle and the distance of the lidar from the ground (described above in detail)
Add PCD files to the tool
Provide the details of the distance from the vehicle front to the target line (Distance from VRP to IRP) and distance from the intersection point to the right corner of the tetrahedron target (Distance from IRP to TRP). The user needs to provide details for all the pcd files.
The target looks like a V shape line with extensions at the end of it in the 2D lidar scan. In order for us to process the calibration, we need sectional information about the face of the target. Hence, we ask the user to paint the four sections as shown below. The painting process has to be done for all pcd files.
This step has the run calibration button. On clicking, calibration runs. This usually gets finished in about 8-10 mins.
Visualisation
Visualisation shows the vehicle axis, lidar axis, the foru target corners and lidar points projected on vehicle space using the extrinsic calibration.
Error Statistics
We present two error statistics: Plane Distance Error and Line Distance Error
Plane distance error: This tells how well the lidar points align with the faces of the model target in vehicle space.
Line distance error: This tells how well the lidar points align with the edges of the model target in vehicle space.
Export
You can export the results from the right corner on top.
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.
vehiclePoint3D is the 3d coordinates of a point in the vehicle coordinate system.
lidarPoint3D is the 3d coordinates of a point in the lidar coordinate system.
