Home/ Questions/Q 7430253
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-29T09:06:32+00:00

I’m using AForgeNet to grab playing cards from a video stream. I grab the

  • 0

I’m using AForgeNet to grab playing cards from a video stream. I grab the whole card and also the section with the actual rank of the card and my problem is that the Template Matching is not really working out for me as its either to sensitive or making to much mistakes.

So I had this idea of splitting the image into sections like this:

enter image description here

If a section has more than 50% black it will represent a 1 otherwise a 0. This will generate a binary representation which I can compare against my “templates”. As its a playing deck its only the chars: AKQJ1098765432 and I think they are unique and few enough to get work. By this it wont matter if the images are 1-2 pixels off.

I do suspect there is something similar like this already that I could reuse, any ideas?

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    I think that a more robust solution is to extract scale and rotation invariant features from the card number and rank. You can try, for example, image moments.

    After you extract those image features, you can train some classifier (e.g. a neural network) to predict the card number and rank.

    • 0