Playing Card Detection: Which techniques to use?

I am working on a project where my aim is to recognize the playing cards in a poker game. I have detected the individual cards using regionprops. Now i have an array which stores each individual card. I am trying to figure out a way to determine these cards correctly. I tried using SIFT with 52 playing cards database, but it failed terribly. Can anyone suggest a better and easier way of doing this? Also i want to know how to separate the cards in the second image(These are the cards of the player).

Answer 1

[Harsha Phadke]

Hey Rohit,

You can use Normalized 2-D cross-correlation to compare the captured images and the images in a 52 playing card database. The value of normalized cross-correlation can act as the threshold for matching the captured images to the images in database. Although, you might need some image pre-processing like rotating the images before you can apply the cross-correlation.

Following link might be helpful:

https://www.mathworks.com/help/images/ref/normxcorr2.html

[ROHIT AILA]

@HarshaPhadke What do you mean by value of normalized cross-correlation? Is there some kind of score to know better matches? When i used normxcorr2, what happens is.. it matches all the suits to something on the card. But i don't know which one is a true match.