Help with number recognition

[AxB]

Hello everyone.

I am trying to use Tesseract-OCR (v3.02) to recognise numbers (typically 5-7 digits). 

The font used is probably "Futura Bold", but styled in a particular manner (see example). 

Using the digits parameter, Tesseract-OCR either gets it completely right, or fails to identify anything at all (return blank) depending on the crop (see example).

Here are two examples where Tesseract-OCR fail:

And is an example where it succeeds:

From all my testing, as long as I do not forget the "digits" parameter, it will get it right if it gets it at all (without that parameter, it occasionally confuses 1 with l or I). And the tighter I make the box, the more likely it will succeed.

The issue is that I am trying to automate the process as much as I can including the image capture step. Since the number can be anything between 5-7 digits I need to a bit of leeway (i.e. capture a slightly larger area).

Perhaps with a bit more trial and error I might find a way to select a constant crop dimension that will work for 5-7 digits (most likely range I need), it would be preferable if I could get Tesseract-OCR to work with the second example.

Is there anything I might be able to try on Tesseract-OCR side of things? I have only just started using this program.

Thanks.