For my course in Art and Media I decided to make an interactive, vision-based system that projects little smurfs on a table. We used a vertically mounted beamer to project the image and a webcam to capture video:
On a technical level, the vision system turned out to be the most complicated part. The intent was to have user proximity and speed determine the behaviour of the smurfs, which turned out to be difficult to implement without a decent visual reference. Both the projection and lighting circumstances interfere heavily with an already grainy image received from the webcam. We decided to simplify matters by making use of colour detection and give the viewers a bright green mitten which was easier to detect.
It turned out ok, and was one of most reliable projects of the course. A little video was made from the system in action (I apologize for this clip, it’s not my fault but the cameraman was pretty crap):
I’m pretty sure it was shitty as a work of art though…