The theory of pattern recognition of simple images
The challenge is to recognize simple objects with a camera phone. Analogues such applications many, many recognize the face, the whole picture and other complex shapes.
Where and how to start from the real-time recognition of simple shapes? Maybe there are some libs frišnyh, docks, manuals?
If the shapes are quite primitive — you can distinguish the outline, to make approximately (all this is easily done using OpenCV, just a couple of lines, + like in the examples there is such a thing) and look at the characteristics of the obtained figures.
Or using Blob Detection — so let's say markers recognize faces (although this is more complicated but again is full of examples)