Golden Eye: Real Time Black Jack Analysis

In this project a machine driven real-time analysis system of the Black Jack game is developed. A mounted table camera provides image data of the game table which is processed with an embedded table computer. Extracted information like visible game cards and chipstacks on the table is combined with information obtained from an electronic chip-tray in order to analyze game flow and to determin the game states. Image processing is separated into the tasks robust card- and chip-detection. In order to reduce the computational costs and therefore to increase the performance the images are separated into regions of interest. Within these regions the objects to be detected are classified by calculating global descriptions. A low-level image processing chain consisting of adaptive segmentation, morphological operations and Hough analysis yields into compactness, area, dimensions and angular distributions. Tests were performed with an experimental setup provided by GRIPS Electronic GmbH, the cooperating industrial partner in this project. Extensive tests proving the robustness of the proposed approach. As the requirement of real-time processing, in this case 4 frames per second, is met by the prototype the system is about to be implemented in the commercial system.