Revision history [back]

First, you will have to calibrate your camera to have the intrinsic parameters (focal length, principal point, distorsion coefficients):

• you can use cv::calibrateCamera
• you will have to take around 10 images like this one: https://github.com/Itseez/opencv/blob/3.1.0/samples/data/left01.jpg
• you can read this tutorial: Camera calibration With OpenCV

Then, you will have to detect the balls. You can look for example to these questions:

• Colour + Contour Detection
• Detecting objects in OpenCV

And finally, you will have to match each ball 3D coordinate to his corresponding 2D image coordinate. The pose can be computed with:

• cv::solvePnP
• Tutorial in Python: Pose Estimation