songyy's profile - activity

Hi there, I'm trying to model the game state of CandyCrush using OpenCV method. My approach is to first detect the candies via template matching, and then compute the look of the game board based.

However, there's one particular problem that.. the candy can be in different states. For example, there can be a "cross" on top of the candy (refer to the image below); and in this case, I found it impossible to matching using my existing candy template.

One way is to take a screen shot of all possible candies, and then further apply the template matching algorithm; but that may be too expensive, because I may have to apply all template through-out the screen.

I therefore wants to ask if there're any idea for this kind of Candy-Matching problem.

Thanks a lot