I’ve spent the afternoon staring at USB traces trying to grok the leap motion at a protocol level, no huge progress (most of the time reminding myself how USB works). I have at least replicated the openleap instructions so can capture raw data.
Here are a couple of scary photos of me in component stereo (simple red/green and rgb with the blue channel is the difference channel) :
The results from openCV’s stereo correlation aren’t quite there yet : (no doubt due to my coding):
The minority report hand wavy stuff is interesting but I’m more interested to see if I can get one working as a well-calibrated robotics image sensor – the field of view on these cameras is immense (150 degrees or thereabouts) and the stereo obviously can’t hurt.
Next stop – fish out an IR laser line from one of the many boxes of bits upstairs (and try not to fry the CCD like I did last time) and do some scanning experiments .