Simplify opencv image writing usage

[CmdrDats]

At the moment, there's quite a lot of mucking around in OpenCV core bits (in particular, drawing onto images)

• Evaluate each usage and see if it's needed
• if it's needed, can we convert to java image manipulation instead (which reduces library usage footprint)

This will be particularly important for later for updating the OpenCV later down the line.

The core usages for openCV should really be:

• Camera feed
• Homography calculation (board to flat 2d projection)
• Reverse Homography calculation (for physical projector, flat 2d projection to board)

There's some experimental work around figuring out the position of the board dynamically to skip or ease calibration , but that is not a primary goal and can be ignored (it's a lot of work for marginal gain).