OCR - array of numbers

[pawhan11]

Hello

I am beginner in image processing and I  need sugestion how to create array of numbers from image.

So far I ve processed this:

to that:

in vision builder.

Now I should divide this square into 9x9 smaller and then run OCR in each of them?

And what about size of the base numbers in OCR, should I create instances of the same number (depending on size?)

Thanks for help

[TiTou]

Good effort!

Can you post your inspection an a few images so we start from a good base to help you?

[pawhan11]

I ve done that in 6 steps starting from:

1) color plane extraction HSL

2)Local Threashold Background Correction  look for dark objects 11x11

3)Binary image conversion

4)Adv Morphology remove border objecst

5)Lookup table Equalize and 4x find straight edge

but sometimes after morphology i get that:

depending on foto quality

thanks for some suggestions

[Simon_T]

Hi!

It looks to me like you have a lightning issue - which is expected since from what I see, there are no controlled lightning conditions in your application.

As the paper folds, the white is different (darker) in the bottom-right corner. This makes tresholding more difficult.

I have a few ideas that you may want to try:

1) Use a flash when taking the picture

2) Try to create two separate phases: separate the paper from the background first, then try to detect the black (ink) on the paper with more strict tresholding settings.

3) Try to set the ROI for the paper , or for the table first (trapezoid), then you can use more strict tresholding without the background interfering.

I hope I helped.

Simon