0

I'm working on a stereo vision system based on openCV which current return correct 3d coordinates, but in the wrong perspective.

I have program a function which give me the camera-3d-coordinate and the expected real-world-coordinate from a cheesboard, but I didn't find out how to generate a transformation matrix from this data.

All possible functions I found in OpenCV doesn't work because they work with 2d-coordinates on a image, and not with the calculated 3d coordinates.

pointhi
  • 303
  • 1
  • 5
  • 13

1 Answers1

0

Check out this answer. It is for 2D points, but if You expand T to 3 + 1 element and make R 3x3, it will work.

Here is code that uses this method.

Community
  • 1
  • 1
morynicz
  • 2,322
  • 2
  • 20
  • 34
  • thanks, I try to understand that, but I don't understand matrix operations yet. I searched through the internet and found informations about Affine Transformation and Homogeneous coordinates which would solve my problem, and your answere look like that. Now I have used `cv::estimateAffine3D` to generate the transformation matrix, but I don't know how to transform my points with that Matrix. The results from `cv::transform` does't look well to me. – pointhi Aug 11 '14 at 13:47