I triangulated the tracked pts from each camera, and now would like to re-project the 3D tracked pts into the image plane of one of the cameras.
I’ve looked both into
opencv.projectPoints and your
camera_calibration.py. I also inspected the
stereo_params.pickle and the intrinsic params of each camera.
I’m debating what are the DLC that I should feed into
I used their
imgpoints_proj_EE, _ = cv.projectPoints(arr_EE, (0,0,0), (0,0,0), \ stereo_dict['camera-1-camera-2']['cameraMatrix1'], \ stereo_dict['camera-1-camera-2']['distCoeffs1'])
(also tried to use
np.zeros((1,5)) as the distortion coeffs but it doesn’t help.
It seems that the tracked points are more or less in the right place, but slightly shifted.
In your camera_calibration.py I saw that for the re-projection error, you used the
tvecs computed from
opencv.calibrateCamera, but that DLC doesn’t save these. Do I need the rotation and translation vectors to re-project? what would you suggest?