What else needs to be done to recognize the image in the image?

Question

C++ / C#

Good afternoon. There is a scene: - a sheet of white paper with text, on this sheet there is a mark (a white circle the size of a 5 ruble coin, with a black outline and a white number 8) and 3 coins (of any denomination).
What I did:
1. Converted the image to grayscale.
2. I did binarization (already at this stage, the text disappeared from the white sheet)
3. I selected the contours with the Laplace transform.
What do I need to do next? Logically, I think you need to do segmentation, it was not possible to do kmeans. The problem is that we have several similar objects, how to mark them up and divide them into different sections for processing, the mark recognition code (that this is the number 8, and not something else) has already been written. Just iterating over the entire image will not work, as there will be many labels in the future.
Please advise how to implement this.