1

I'm trying to calculate the camera pose from a "warpPerspective" processed image. I start with the original camera image, then I undistort it and finally get the bird view one:

enter image description here

I've already estimated the camera pose from both the original and the undistorted images using solvePnp() and Rodrigues() where I've passed the instrinsic matrix and distortion coefficients taken from the calibration process (for the undistorted case I had to use cv.getOptimalNewCameraMatrix instead of mtx and 0 for dist in solvePnp(), more details here in comments section).

Now I have the image points(2d) and the corrisponding real world ones(3d) for the bird view image but I don't know what camera matrix I need to use as input for solvePnp in this case (I assume that the distortion coefficients must be 0 as the "undistorted" case, right?). Any ideas?

Marco Ghigo
  • 47
  • 8
  • Unable to help you with your problem, sorry. But the second cow/python/openCV question today ? What are the chances ? :D https://stackoverflow.com/questions/74595357/how-do-i-segment-head-and-body-in-a-top-down-depth-map-of-a-cow#comment131674361_74595357 – nick Nov 28 '22 at 14:59

0 Answers0