This little project is an basic augmented reality system made with openCv library (algorithms to grab the code from image
and get a frame from webcam) and irrlicht 3D (render 3D objects over the webcam scene). The main propose of
this project was the implementation of an artificial neural network (in that case used to recognize the code in AR system).
Programming language: C++
IDE: Code::Blocks
3D engine: Irrlicht
Other: openCv library (http://sourceforge.net/projects/opencvlibrary/)
There are a few steps to render the scene:
-> Using openCv library we get a frame from webcam.
-> Opencv also provides an algorithm to get squares in the image, we get the most evident
square (this step is not goog, too slow, i think we can get more speed using openCL to paralellize the algorithm).
-> Now we send the code (square at upper-left corner of the screen) to an Neural Network (Neural network is not
the best way to recognize the symbol, but my teacher wanted it. The result of the network determines the 3D object.
-> Using an conversion function we get the image data from opencv frame and write into an irrlicht texture.
-> Now we have the webcam image on an a irrlicht texture that can be rendered on the screen with the 3D mesh.
conclusion: too much work for nothing.. but it was an interesting experience.