Object Extraction and Recognition in Simulink

I'm constructing a model to recognize and detect suits of cards in Simulink. The suits are on white signs and the background is not solid (live video outdoors). I am trying to use thresholding and blob analysis then determine which suit is which. I've found ways to somewhat reliably distinguish between each suits if I can extract it from the rest of the picture.

Diamonds and Hearts seem to work well by thresholding the Cr value (not much red in our background). I run into problems when trying to determine Clubs and Spades. The best I've been doing is thresholding the ' values of Y'CbCr. It always gives me blobs of black, but I'm also left with the dark green color of the trees. Here is a color and binary image. As you can see the dark green color from trees is included in the binary image. I've also tried thresholding for white values (the signs), but I often get the trees as well.

Could anyone offer some advice or insight as to how they would attempt to extract the desired features? Thanks.

Answer 1

[Nicu Chira]

Hi, i'm a beginner in matlab but i have an idea, can you use?, several auttothresholding, and each is for a specific desired color, and then combine theirs output singnals (image), and the white should be only for the chosen colors to be "1" (boolean), like a color filter i think.

Also, i have question, i have a project in matlab simulink and i have to calculate the distance between the camera and the object. I used "Tracking an Object Using Correlation" from matlab exemples, but the problem is that, if i move to much the object closer or further it loses it, and i want to extract only the image of the found obeject, to calculate the distance, based on the number of pixels it has. Well can you or anyone elese, who reeds this topic, can give me an idea? Thanks.