Skip to content

Calibration

Rui Miguel Carvalho edited this page Sep 14, 2020 · 1 revision

Realsense tools

Intel Realsense D435i do not come calibrated from factory so it is necessary before running any VIO algorithm, some of these like ROVIO and VINS-FUSION can perform online calibration. Nevertheless, it is very important to provide an educated initial guess.

Luckily, Intel has provided some calibration tools recently that we can use:

For that launch the realsense-viewer with:

roslaunch realsense2_camera realsense-viewer
  1. On chip self calibration - please follow this video tutorial.
  2. IMU initial calibration - run the python script according to the the official guide and follow the on-screen instructions.

Thirdparty Open-Source tools (Kalibr and IMU utils)

For more info on how to perform the calibration follow this tutorial (it's in chinese so you probably want to use a translator...). Also make sure to check the official docs for more info on what frequency to run the camera streams (you can modify this by using the ros package [topic_tools] (http://wiki.ros.org/topic_tools).

Make sure the .yaml target file matches the actual calibration target. Calibration files in this repo were obtained with an A4 printed april grid.

After obtaining the .yaml file from doing both the in/extrinsics and IMU calibration, additional scripts can be found under /work/kalibr_ws/kalibr/aslam_offline_calibration/kalibr/python/exporters/ to export to formats suitable for ROVIO(LI). Regarding vins and ORB-SLAM config files I couldn't find any automatic tools to do it but it is just a different way to express the obtained parameters and should be fairly straightforward to do.

Notes

  • The IMU utils submodule repo was modified to avoid common issues with dependencies, please refer to this issue.

  • The IMU utils contains some .m files to plot the allan curve of the IMU sensors. These can be plotted using Octave instead of MATLAB without any additional package/toolbox.