RQDecomp3x3 returns wrongly signed euler y-angle (Feature #743)

Added by Tobias Langner over 14 years ago. Updated almost 14 years ago.

Status:Done Start date:
Priority:Blocker Due date:
Assignee:Vadim Pisarevsky % Done:

0%
Category:calibration, 3d
Target version:-
Difficulty: Pull request:

Description

The corrected line should be:
eulerAngles->y = acos(_Qyr0r0) * (_Qyr2r0 >= 0 ? 1 : -1) * (180.0 / CV_PI);

I'm still a bit confused why all your rotation matrices are transposed.

calibration.diff - patch for RQDecomp3x3 (1.2 kB) Sven Lange, 2011-04-28 09:49 am

calibration.2.diff - patch for RQDecomp3x3 (1.2 kB) Sven Lange, 2011-04-28 09:49 am

Associated revisions

Revision 6c58a22e
Added by Andrey Kamaev almost 12 years ago

Merge pull request #743 from jet47:fix-sparsemat-iterator

History

#1
Updated by Sven Lange almost 14 years ago

I think tlangner is right. I also think, that the bug should be fixed very soon, because currently the function is useless for calculating euler angles.

I did some testing and compared the results with the analogous matlab function. The result was always a wrong signed rotation around the y axis. Values are correctly.

The reason for the wrong calculation has its origin within the rotation matrices. If you compare them with the matrices in “Multiple View Geometry (Hartley and Zisserman)” p. 579, you will see that Qx and Qz are the transposed matrices and Qy is not transposed. So Qy should be [c 0 -s; 0 1 0; s 0 c].

As a result the index for calculation of the sign for the y rotation is wrong. It should use index 3,1.


  

  
      

        
#2

        Updated by Vadim Pisarevsky almost 14 years ago
        
      


thanks! the patch has been applied in r4965
  • Status changed from Open to Done
    •  
  • (deleted custom field) set to fixed

  














  

    











  









Also available in:
	Atom
	PDF