1
$\begingroup$

Pose Estimation of single camera and planar pattern:

If I have an appropriate camera matrix with intrinsic parameters that allows me to calculate translation of the camera relative to a chessboard, for example, and it does so very accurately, does that mean my rotation calculations will also be accurate?

This may also be considered in the context of the camera calibration documentation for opencv.

References to your answers would also be appreciated.

EDIT: thanks to @tbirdal for the links "Pose and position" is an incorrect phrase. Pose is a combination of position and orientation.

$\endgroup$
  • $\begingroup$ Please elaborate. Camera matrix doesn't represent your camera position but rather describes how it perceives the real world: projection. So in what context are you referring to that? Do you have a stereo or monocular setup? $\endgroup$ – Tolga Birdal Dec 7 '14 at 0:33
  • $\begingroup$ @tbirdal, Edited question $\endgroup$ – Grim Dec 7 '14 at 0:40
1
$\begingroup$

If you pose estimation algorithm carefully minimizes the reprojection error using the standard non-linear techniques e.g., Levenberg Mardquardt, and your corner detection scheme is sufficiently accurate in the subpixel level, then yes. I would expect the residuals to be very small. Of course take into account the distortion parameters; they have to be calibrated as well. If you are using OpenCv and referring to the standard functions such as findExtrinsicCameraParam, then you are good to go.

The problem is also known as PnP due to the correspondences. Lepetit et. al. review it nicely here.

Also, Grest et.al provide an overview of 3 methods here.

$\endgroup$
  • $\begingroup$ Thank you for the links, they proved very informative. Question was re-edited. Grest et.al showed that translation and rotation are very closely correlated when doing the estimations with all three methods. However when there is a deviation between translation and rotation, then translation would become erroneously estimated more easily than rotation. This answers my question. $\endgroup$ – Grim Dec 7 '14 at 1:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.