When I’m looking at OpenCV: Basic concepts of the homography explained with code，I noticed that there seems to be an error here.
In the part of Demo 3: Homography from the camera displacement
I think the following explanation is not very appropriate：
c1Mo is the camera pose for the camera 1
c2Mo is the camera pose for the camera 2
According to the above derivation, cMo is a matrix that converts object coordinates into camera coordinates, therefore cMo is the position of the object plane in the camera coordinate system.
Therefore, I believe the above statement should be:
c1Mo is the object plane pose for the camera 1
c2Mo is the object plane pose for the camera 2
I don’t know if I misunderstood，the original text expressed the same meaning as what I understood？Or if the original question was expressed incorrectly？
you are right to point to the article’s… mistakes/bad writing. the article’s wording is the issue.
the matrices we’re dealing with when estimating poses are always matrices like “cMo”, where “o” is object. the common frame is the camera frame. the object’s pose is estimated, in the camera frame.
having two cameras/view, one still has “c1 M o” and “c2 M o”, presumably for the same object.
those are not camera poses, as written in the article, but object poses. it’s just that they’re poses for the same object, but different views.
all the actual math in your screenshot looks okay to me. I did not check the last part of the last matrix that involves “-R t”.
please submit an issue on opencv’s github.
if you want to, you can edit the source of the article at opencv/doc/tutorials/features2d/homography/homography.markdown at 4.x · opencv/opencv · GitHub and submit a pull request.
