CodeProject.ai with License Plate Reader module - HowTo?

An enlargement of the plate is going to look granular because you're enlarging just a small part of the picture. Effectively you're digital zooming for which the tehnical explanation is you're reducing the pixel density and thus the definition as there are less pixels per square inch. Try focusing in on the small part of any photo on your pc and the more you enlarge it, the more blurry and more granular it becomes. The results you have look pretty typical to me for an enlarged plate.

There are only 2 ways you'll get a less granular picture:

1. is to set up a camera soley for LPR and zoom it in more optically (use a variable focal length camera so it can be zoomed in then fixed), so it captures an even tighter licence plate shot at the expense of the rest of the vehicle. Use a 2nd camera for over view. That said, you camera is pretty tight anyway.

2. The other is to use a higher resolution camera, although this may cause issues at night with shutter speed. Usually less of a problem with a dedicated LPR cam as most people switch them to IR mode at night and use supplemental IR thus effectively increasing the amount of available (invisible to the eye) light. However don't chase pixels. Higher res will result in less light being gathered unless there's a corresponding increase in sensor size / sensitvity or the addition of more artifical light.

LPR is a science and I'll let others address the details as it's outside of my scope of experience. But above are 2 general observations.
 
Is there a way to define regions on ALPR (e.g. pt, es, uk,...) to optimize the recognition of plates to a specific country, in my case Portugal?
I'm having issues with some letters (e.g. 0 and O, ...)