OCR seems like a more practical purpose than machine learning for the task

Yea this seems like a straight forward OCR problem (I'm not sure what openly available tools exist for that at the moment but I'm sure there is something)