I can work around the issues that kalibr_calibrate_cameras raise. Also, I should learn more about ROS.
Until now, I cannot run kalibr_calibrate_cameras, because „No corners could be extracted for camera /cam0/image_raw”. I thought that the four parameters needed by Kalibr to calibrate a camera – fu, fv, pu and pv – were not the same as the ones given by OpenCV, but I was wrong. Those parameters are described in the 3×3 camera_matrx, vectorized as [fu, 0, pu, 0, fv, pv, 0, 0, 1].
8 September note: No, you still need this part to work. The camera needs to localize itself w.r.t. a target, which is the same calibration board.
I want to create a ROS bag with camera data and IMU data. Previously, to make the dataset for kalibr_calibrate_cameras, I used my own image_grabber.cpp and assumed similar update frequencies of the two cameras. For the combined bag, I should use some tools that can do that already. ROS should have those tools. I went through their tutorial, and now will look into the image_pipeline and xsens_driver.