Revision history [back]

Hi All

I am attempting to get the ideal pinhole camera for the camera of an iPad Mini 4. I have done multiple calibration sessions using a chessboard pattern and code from: https://docs.opencv.org/master/dc/dbb/tutorial_py_calibration.html

The issue I have is the consistency. I have used up to 200 images per session. The checkerboard is detected in all of them. The resolution of the images is 3024 × 4032

I find the deviation rather large, for the same camera.

fov_deg    59.2862
fov_deg    60.6048
fov_deg    59.9288

f_x    2913.33
f_x    2907.07
f_x    2933.18 

f_y    2914.59
f_y    2901.48
f_y    2930.74

c_x    1238.07
c_x    1282.58
c_x    1275.43

c_y    1605.38
c_y    1568.35
c_y    1574.36

Is this expected? Which accuracy is realistic? Am I missing something?

This data is later used for 3D reconstruction, so it should be as accurate as possible.

Looking forward to your insights

Best regards Wim

Hi All

I am attempting to get the ideal pinhole camera for the camera of an iPad Mini 4. I have done multiple calibration sessions using a chessboard pattern and code from: https://docs.opencv.org/master/dc/dbb/tutorial_py_calibration.html

The issue I have is the consistency. I have used up to 200 images per session. The checkerboard is detected in all of them. The resolution of the images is 3024 × 40323024 × 4032 Auto-focus is turned off.

I find the deviation rather large, for the same camera.

fov_deg    59.2862
fov_deg    60.6048
fov_deg    59.9288

f_x    2913.33
f_x    2907.07
f_x    2933.18 

f_y    2914.59
f_y    2901.48
f_y    2930.74

c_x    1238.07
c_x    1282.58
c_x    1275.43

c_y    1605.38
c_y    1568.35
c_y    1574.36

Is this expected? Which accuracy is realistic? Am I missing something?

This data is later used for 3D reconstruction, so it should be as accurate as possible.

Looking forward to your insights

Best regards Wim

Hi All

I am attempting to get the ideal pinhole camera for the camera of an iPad Mini 4. I have done multiple calibration sessions using a chessboard pattern and code from: https://docs.opencv.org/master/dc/dbb/tutorial_py_calibration.html

The issue I have is the consistency. I have used up to 200 images per session. The checkerboard is detected in all of them. The resolution of the images is 3024 × 4032 2448 × 3264 Auto-focus is turned off.

I find the deviation rather large, for the same camera.

fov_deg    59.2862
fov_deg    60.6048
fov_deg    59.9288

f_x    2913.33
f_x    2907.07
f_x    2933.18 

f_y    2914.59
f_y    2901.48
f_y    2930.74

c_x    1238.07
c_x    1282.58
c_x    1275.43

c_y    1605.38
c_y    1568.35
c_y    1574.36

Is this expected? Which accuracy is realistic? Am I missing something?

This data is later used for 3D reconstruction, so it should be as accurate as possible.

Looking forward to your insights

Best regards Wim

Edit: added info about auto-focusauto-focus Edit2: corrected image resolution