The potential of solving real-time demanding industrial applications, using vision-based
algorithms, drastically grew due to an increasing availability of computational power.
In this thesis a novel real-time, vision-based BlackJack analysis system is presented. The
embedding of the vision algorithms in a compound system of other information sources such
as an electronic chip tray, reduces the vision task to detect cards and chips. Robust results
are achieved by not just analyzing single frames but an image stream regarding game-ß ow
informations. The requirements for such a system are a highly robust and adaptive behav-
ior. This is motivated by the vital interest of casino entrepreneurs in a 100 statistical
analysis of their offered gambling in order to support the business plan, measuring table
and dealer performance and give accurate player rating. Extensive experiments show the
robustness and applicability of the proposed system.
