How should I approach identifying multiple instances of specific shapes

I would like to identify Set cards from an image using OpenCV 2.3. I can successfully use findContours to isolate the shapes:

How should 开发者_如何学GoI encode the three shapes of interest and compare them to the image so that I can find them in different rotations and sizes?

I'm not asking for a solution, just suggestions on an approach to take.

From the left image, you can segment each card out.

Then try to use OpenCV matchShapes() to match these three shapes.

As matchShape() uses HuMoments, opencv doc described that Hu Moments are invariants to the image scale, rotation, and reflection except the seventh one...

You can also reference opencv's example code: cpp/squares.cpp to design your own shape matching program.