Closed jbohren closed 9 years ago
+1
I would vote for m
I was also surprised by this, but the ros stereo calibration appeared to generate mm values when I ran it.
Upon further reflection it looks like this is actually pixel units, in which case it should just be (Tx/fx)
which in my case is Tx/1013.865828
. Then this should give the offset in meters.
I hate that the projection matrix handling seems to have to be reimplemented in every program that listens to CameraInfo.
According to https://github.com/ros-perception/image_pipeline/blob/indigo/camera_calibration/nodes/cameracalibrator.py#L383 when calling the cameracalibrator.py the size of the chessboard square should be provided in meters and then the result of camera calibration parameters will be in meters, too.
Just a remark, the correct formula is -Tx/fx.
Just a remark, the correct formula is -Tx/fx
Right, that's what I said above. I'll update the PR later today.
Great, thanks Jonathan for your contribution!
Ok the fix is in there.
+1
As an aside, this works fantastically with the VISP implementation of Tsai's hand-eye calibration algorithm: http://wiki.ros.org/visp_hand2eye_calibration
It's good to know about it! Thanks :+1: !
:+1: