Sorry if this is a recurring question, but I can't find the right keywords for this search.
I need to develop a system for visual recognition of labels attached to products from a warehouse. I'm using a fixed focus camera, so the idea is to use a label with some code with 6 alphanumeric digits printed in a large font. Then this system would be responsible for performing ROI extraction and applying OCR to recognize the objects in the scene.
My main problem is the ROI extraction part. I tried to use template matching, but due to the difficulties with scale and rotation, it doesn't seem to be the right technique for the application. I also tried to use feature matching, but the results are still insufficient.
My question is how could I develop the label to facilitate ROI extraction? Could I use something like apriltags to simplify the homography?
Thanks in advance!
