Particular camera setup and calibration process with opencv

Question

This is my first post and am delighted to be a part of this community.

I've two webcams and want to use it to detect 2d coordinates of the center of the white tennis balls. And find the 3d coordinate of the centers. My cameras are setup like this. There's a circular region of radius 7 feet and the cameras are placed on opposite end of each other which means if camera1 is placed at 0 degree of the circular region then camera2 is placed at 180 degree of that same region so they are on a straight line exactly opposite of each other on the circumference of the circular region.

I need to calibrate the cameras and need to find intrinsic and extrinsic parameters.I'm using opencv for this.

Can I use cvStereoCalibrate() for this camera setup?

I'm asking this because if you look at the camera setup you will see that there's a point on camera1 and camera2's captured image that is collinear with epipoles of both cameras. So the epipolar line is a point. Will this be an issue for calibration procedure? If yes is it possible to tell me how to solve this problem?

Answer 1

This is a camera setup with the stereo cameras toed-in completely (really cross-eyed :). To calibrate you need correspondence points from each image. You can only find these around the edges ... all the way around in the peripheral vision of each camera and even then I think they would be mirrored from what the algorithm would expect (have to think about that one). I have not tried this, but I think this would warp the image into a doughnut assuming it didn't completely confuse the algorithm beyond recognition. The stereo correspondence algorithms would definitely have problems with this, I think, Because also far in one is near in the other.