Monday, April 27, 2009

DiamondTouch

I was trying to get a Mac program which would talk to the Diamond Touch machine with Jeff today. Jeff ended up complaining that the final result was far less accurate then the numbers he was used to ( 0-171 rather than 0-3000ish).

I suspect that they are doing some sub pixel reconstruction but why convert back to a 0-3000 odd value why not 0-1.0 (i.e a float). I'm not sure why perhaps I'm not right. Jeff seems to think that they have 3000 odd wires in the board (I'm not convinced).

I found this paper which looks easy to implement. Then I need to scale back to 0-3000 rather than use 123.434 pixels

http://www.iop.org/EJ/article/1464-4258/3/4/308/oa1408.pdf?request-id=28089999-5a85-47e4-b661-e7226eef9607

Perhaps I'm wrong and they do have 3000 pixels and for some reason the board is only reporting it has 171 wires(samples). In which case the DT people are missing something.

No comments: