As you know if you're following this blog, we in the projector group have finally started working with an actual projector. We've only been able to test it with the projector aimed at a wall (since we haven't gotten it mounted yet), but it's still been very useful.
The most important discovery we made was that our initial idea to calibrate the video stream by finding a colored rectangle doesn't work as well as we'd hoped: the hue of the rectangle in the browser window is radically different from how the camera sees the projected image; red becomes purple and yellow becomes green. We have a few ideas of how we could tackle this problem, but we decided to simply change method instead.
We tweaked our existing code to only use AR markers for the calibration. This is a method we've had in mind for a while (kind of a plan B), and it turned out to be a better solution: With our first method, the program could find several colored rectangles with no way off knowing which was the right one. An AR marker represents a number between 0 and 1023, removing all risks of ambiguity. Hopefully, we'll still find some use for our color filtering code, since it was pretty cool : )
In conclusion, I think you could say that we've taken one step back and two big steps forward. We'll soon post info on our plans for the coming weeks, together with a few storyboards describing our new method.
No comments:
Post a Comment