ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
1

velodyne calibration

asked 2013-10-10 22:43:22 -0500

sai gravatar image

ubuntu 12.04 and ROS Fuerte

Hi,

The velodyne_pointcloud package runs fine and display pointcloud from velodyne 32e using the calibration file that comes with the ROS package.

Questions:

  1. I was wondering if one can use the specific calibration file that ships with velodyne instead of the one that comes with ROS package?
  2. The format of the calibration files is different and their values change a lot.Can any one tell how to use the specific calibration file rather than the generic calibration file?

Snippet from the ROS calibration package

{dist_correction: 0.0, dist_correction_x: 0.0, dist_correction_y: 0.0, focal_distance: 0.0, focal_slope: 0.0, horiz_offset_correction: 0.0, laser_id: 0, rot_correction: 0.0, vert_correction: -0.5352924815866609, vert_offset_correction: 0.0}

32db.xml that comes with velodyne can be found at this link. https://dl.dropboxusercontent.com/u/95042389/32db.xml

3.Does using the calibration files improve the accuracy? If possible can tell quantitatively?

4.As kinect has a camera model that can be used to relate camera parameters and real world points, Does velodyne has some model. Can you point some source for the geometric or mathematical model of velodyne 32e?

Thank you so much for the help.

Sai

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
1

answered 2013-10-11 07:23:43 -0500

joq gravatar image

Velodyne does not provide identically formatted information for all their models. Mostly for that reason, the driver design includes translating the information it needs into a common YAML format. See the design review discussions for details.

The gen_calibration.py script will generate the desired YAML file from your device-specific db.xml.

The params/32db.yaml included with the driver was generated in that way from an example 32E device. Several other 32E devices seemed to share the same calibration. I don't know if they are all the same, or just some of them. It seems reasonable to assume that using your specific calibration will improve accuracy, or at least not hurt. Please try it and report your results here.

The best way to measure accuracy I know is to find some features in the real world that stand out clearly, physically measure their distances, and compare with the generated point cloud. Rviz has a tool to determine the coordinates of specific points in a cloud.

The Velodyne has no RGB camera like that on the Kinect, so camera calibration does not apply.

edit flag offensive delete link more

Comments

Thanks @joq ! I generated the specific calibration file and tested the velodyne with both calibration files with a known landmark position. Surprisingly, both give same 3D positions of known landmarks(accurate to centimetre).

sai gravatar image sai  ( 2013-10-11 20:38:53 -0500 )edit

Do you have any idea about the errors in measurements from velodyne? Do they vary as distance increases ? Any way to model them or the parameters they depend on ?

sai gravatar image sai  ( 2013-10-11 20:39:28 -0500 )edit
1

I would expect some quadratic effect on the error, but have not measured it. You can try measuring features at different distances.

joq gravatar image joq  ( 2013-10-12 05:43:20 -0500 )edit

Question Tools

2 followers

Stats

Asked: 2013-10-10 22:43:22 -0500

Seen: 3,598 times

Last updated: Oct 11 '13