Has anyone been successful in developing a robust OCR application capable of reading multiple fonts?

[b.ploetzeneder]

Hi,

I've done a number of food label projects (I used to work in food & packaging).

From my experience, you do OCR to read "dynamically" printed stuff like dates, weights etc, and that is usually easily implemented. I never had trouble with that in LabVIEW.

To identify which product / packaging you are dealing with, OCR is not really suitable, often you do just pattern matching on some label.. Basically, a database with 1-more identifier per label, and from there on you figure out where to go. OCR fails all the time with that, because you have graphics in the background, fonts change, font sizes change.. in general food packaging designers are a bunch of people who dont care about process automation.. I recently wrote an article on PM in print inspection.. here:

http://www.vision-systems.com/content/dam/VSD/solutionsinvision/Resources/Microsoft%20Word%20-%20art...

Maybe that helps

Birgit

[UMASO]

b.ploetzeneder 

I like your approach of climbing back to customers' or application/system requirements, before sticking to one method of development.  

[SamiF]

Your article seems unfortunately not available anymore as the link returns an error 404 (page not found).

Sami