OCR of playing cards in a video stream

I investigating if there is a not to complicated project to have a camera mounted above a bridge table and identify played cards from the feed. A lot of bridge tournaments are broadcasted on-line but it demands a person to sit wi开发者_如何学Goth a laptop and click ever card being played and is a tedious job :)

There are 4 players and my thought was to have a marked area for card to be played in before registered.

A few things I'm thinking of.

By using a marked for played card OCR has only have to be made in few occasions but can I reach a 100% success rate? Will I need some uber machine to handle the OCR calculations and is there fast enough routines for what I want done?

Would be nice to hear your input, suggestions and if you have any experience and ideas!

You may use an Image Correlation. It's fast and precise enough.

Example in Mathematica identifying the suit (the same can be done for the value):

Will something like this work? https://vimeo.com/3563858