openCV shape recognition help

jedbar's picture

Hi there,

I’m an industrial design student (definitely not a programmer) looking at making a learning tool -and I really need some help!

I want a user to be able to construct a basic 2d shape out of popsicle sticks and an overhead camera to recognize the shape and play an audio track which correlates.... ie, the user makes a triangle out of sticks and the computer plays “well done, it’s a triangle”. So a basic requirement, but quite difficult to pull off.

I think CV is the best solution, as I want the kids to be able to construct basic shapes by themselves (and not on a grid), and the linear construction elements (popsicle sticks) should be disposable - so sensors attached to them aren’t really appropriate) That said, IR leds would be ok

I’ve been considering openCV with a database of comparative images of shapes (haar learning) but it’s beyond my knowledge. I do however have some experience with quartz composer, so that would be my first choice.

Any suggestions for setting up what is basically computer vision pattern recognition?

cheers, I really appreciate your time,

Jeremy