Python: Fish-eye PnP Distortion Coefficients

asked 2016-12-09 06:35:05 -0500

updated 2016-12-10 22:34:53 -0500

I have the intrinsic parameters of fish-eye camera with fisheye model in opencv310 and I want to calculate the rotation vector and translation vector for arbitrary image using pose-from-N-points. As far as I know, the Python function, FindExtrinsicCameraParams2, could work, but it would require distortion coefficients p1 and p2 which are not returned from fish-eye calibration. What can I use for fish-eye p1 and p2 and what do they model?

edit retag flag offensive close merge delete

Comments

1

Do you have two cameras? I'm not sure it makes sense to have a rotation and translation "for any image". From a single camera with known 3D world to 2D image point correspondences you can compute a rotation and translation with pose-from-n-points, but that doesn't seem to be what you are asking for.

Der Luftmensch gravatar imageDer Luftmensch ( 2016-12-09 07:07:54 -0500 )edit

Thanks for your reply. Your idea may be not what I want to ask for. The function FindExtrinsicCameraParams2() can finish this work that I want. This function is just fit for Zhang's method but not fish-eye model in Opencv3.1.

damingxing gravatar imagedamingxing ( 2016-12-09 20:15:51 -0500 )edit

Please link to the documentation of the OpenCV function you are referring to.

Der Luftmensch gravatar imageDer Luftmensch ( 2016-12-10 11:11:01 -0500 )edit

Thanks for your reply. link text Here is the documentation of this function. You can find it under the 13th function, solvePnP, in this webpage. You can also note that the 4th parameter, distortion_coeffs, is [k1,k2,p1,p2], which is just fit for Zhang's method. The fish-eye model I used is following, link text. Thank you very much.

damingxing gravatar imagedamingxing ( 2016-12-10 18:32:26 -0500 )edit

I updated your question with many changes. Please be clearer the next time you ask a question. You must give much more context.

Der Luftmensch gravatar imageDer Luftmensch ( 2016-12-10 22:35:59 -0500 )edit