Hello,

I'm working on a project which involves detection of 2D surfaces and the text written on them using a webcam. The surface is being detected and I've been successful in enclosing it in a bounding box.

Now, for text detection I thought if I can crop the snapshot from the webcam based on the region within the rectangle (the bounding box), maybe I can run OCR (optical character recognition) on the cropped snapshot.

But I'm neither able to crop it or recognize the printed characters in any way. I tried running OCR and MSER method already but in vain. I want to avoid neural networks as much as possible as I'm not good at it. Is there a simple way to do it?

I'd be grateful if someone can help me as I'm running late on the project submission!