I have been developing a game for a while now for an outdoor screen instalment, where people can interact with a built in webcam that tilts downwards.
The game is simple really, balls fall from the top of the screen and people within the square must run around to try and pop them! It works using the OpenCV addon, utilizing background subtraction and then I create shapes from each of the blobs detected using Box2D which are used for collision.
The problem I am having is, if people are already in the area in front of the screen when the background image is taken for reference, it obviously messes up when those people move. I need the area to be empty first, but I can’t guarantee that this will be the case, so I am looking for some alternatives.
I have been experimenting a lot, trying not to use a background reference image for blob detection but it hasn’t gone well at all. Anybody got any ideas of an alternative, or any tips/algorithms/techniques that make for a more precise blog detection please?
Any tips would be greatly appreciated.