Image-based dice interpretation for a Yahtzee game

In this paper we present our robot for image-based dice interpretation for a Yahtzee game. The system was developed by a project team of high potential high school students interested in engineering and computer science and won a competition for such projects organised at the University of Siegen in Germany. Within the project, a hardware prototype for a Yahtzee game was designed and created. After throwing the dice mechanically, an image of the scene is taken by a camera. Subsequently, a robust image interpretation software analyses the scene counting the dice dots fully automatically. This is definitely the most innovative and important module of our system. Finally, a software module for mathematical reasoning gives hints for further optimal behaviour in terms of the Yahtzee game. The comprehensive evaluation of our system shows its high robustness against changing environment conditions, especially for the image interpretation module.