you can use color detection + Kalman filter
- transform the video frames into HSV color space using cvtColor
- find a lower and upper bound of HSV values of the color of your balls
- use inRange to detect the ball, this gives you a binary image where the ones are the pixels of the ball
- then you may calculate the average of positions of ball pixels to find the center position of ball
you can then feed the position of ball into Kalman filter tracker.
