calculate stereo calibration

[chomamia]

Thank you for your wonderful contributions. I have a question: I want to calculate independent stereo calibration for each camera (the chessboard images are not captured simultaneously from both cameras, and the distance used with the chessboard between the two cameras may differ). How does this affect the results?

[sebastiengilbert73]

Hello chomamia, If the global reference frame is the same, meaning the 3D coordinates of the chessboard feature points are consistent, it should be fine. Note that if one camera is further away from the chessboard, its calibration quality will be degraded with respect to the camera that is close to the chessboard, since the angular variation between two neighbour feature points will be smaller.

[chomamia]

Thanks for your reply. I have a question, I want to use this system on a tennis court, I want 2 cameras that are as far apart as the width of the court with a certain tilt to be able to see the whole court. The coordinate corner is in the middle of the net. After calibrating the camera, for points close to the camera, there is a small deviation, the further towards the end of the court when the distance is farther, the deviation can be up to 2-3m. Can you suggest how to improve it better? I thank you very much for that.

[sebastiengilbert73]

Hello chomamia, Maybe you could test if this error grows with the radial distance from the image center. If it is the case, the problem could be related to the radial distortion of your cameras. As a side note, in an application like this one where the object of interest moves very fast, you'll have to make sure that both cameras are synchronized. Otherwise, the object could have moved significantly between frames from the left and the right camera.