calibration problem

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

calibration problem

mgfleming53202
I'm trying to calibrate a single camera, for use in a shape from
motion problem. The camera is a Nikon D300 with 60 mm Nikon macro
lens. I took 10 pictures of a chessboard, and ran them through
cvFindChessboardCorners and cvFindCornerSubPix. I confirmed that all
of the corners were identified correctly in each of the 10 images. I
then used cvCalibrateCamera2 to obtain the intrinsics and distortion
coefficients, which were then used to try to undistort a series of 6
images (taken with the same camera and lens, of course), starting with
cvInitUndistortMap. The undistortion does not work, and when I look at
the first few values of mapx [obtained from
cvInitUndistortMap(intrinsic,distortion,mapx,mapy)], they seem to be
gibberish (90.0, -118.0, -23.0, etc). I was wondering if someone with
some experience could have a look at my intrinsic and distortion
values and tell me where I might have gone wrong, as I'm not really
sure what these are supposed to look like. They are:

<Intrinsics type_id="opencv-matrix">
  <rows>3</rows>
  <cols>3</cols>
  <dt>f</dt>
  <data>
    1.14660566e+04 0. 2.04579346e+03 0. 1.14588232e+04 1.51486438e+03 0.
    0. 1.</data></Intrinsics>

and

<Distortion type_id="opencv-matrix">
  <rows>5</rows>
  <cols>1</cols>
  <dt>f</dt>
  <data>
    0.41165957 -31.90976715 3.71374376e-03 -3.90885631e-03
744.66357422</data></Distortion>

To me the intrinsics look OK but I'm less enthusiastic about the
distortion.

I have the O'Reilly book but still might benefit from a working
calibration example, with both code and the actual images. I noticed
someone else recently asked for the same.

Any assistance would be greatly appreciated.

Matthew Fleming, MD
Dept of Dermatology
Medical College of Wisconsin