Puzzle games (like Rubik’s cube) solver

This project is a continuation of the PTAM (Parallel Tracking and Mapping for Small AR Workspaces) like approach solution [1]. Basically, it turned to be not an appropriate solution for recognizing full cube under not ideal illumination. Since the process might fail during the processing of any face out of six, it has to start from scratch, thus it will not be finished at all. On the contrary, this algorithm gave a better understanding of Tracking and Mapping problem as well as Structure from Motion problem. At the current stage it was decided to use approaches from Sudoku Solver and First version of Rubik’s cube face extraction to reconstruct cubes model. At the same time PTAM can be used to track cube motion but not rotation. It was also decided to use Kalman filter to optimize PTAM work and to use it for motion prediction because of its nature. This project also contains self-written Rubik’s cube 3D model rendering with the newbie solution method programmed to solve the cube in real time. This means that program can reconstruct cube from the video, not ideally, it might take some time and a lot of rotations until the cube is recognized, and solve it while displaying all rotations on the 3D model and writing all movements as a text to the text box.