JavaANPR training for different format of plates

Question

I found the JavaANPR library and I want to customize it to read my country's license plates.

It seems that the included alphabet is different than what we use ( http://en.wikipedia.org/wiki/FE-Schrift )

Here is an example of a license plate in my country (http://en.wikipedia.org/wiki/Vehicle_registration_plates_of_Malta)

I've already modified syntax.xml to include a definition of our plates' syntax (three letters followed by three numbers)

However documentation for the library is quite lacking and I haven't really found a way to contact the author.

From what I can see I'd have to create a new alphabet with the typeface used on our plates, and train the neural network for it. I tried to do so (as best as I could understand) but it still failed to recognize our plates.

Does anyone who has used the library know what the process of localizing it entails? I need to know how to create a new alphabet and how to correctly train it. I'm not sure if there's anything I need to change in the config.xml file as well.

Thanks

Answer 1

Check out OpenALPR (http://www.openalpr.com). It seems to be more accurate than JavaANPR and may do better for your countries plates. Those Malta characters look similar to the German plate font.