I put together a small script using *nix commands, which can also be run under Windows if it has the correct packages installed, that converts the first 20 pages of a pdf to text and then searches for ISBN-13:, then ISBN-10: and lastly for ISBN:.
It will then rename the pdf with the ISBN number found and move it to another folder. I have Calibre setup so that it reads the isbn number from the filename and then I only have to select the imported books and download metadata.
For books that don't have an ISNB number or in the wrong format, I don't modify those.
Sofar I got a hit rate of about 90%, mostly on technical pdf's.
If anyone is interested I'm willing to write up what is needed in the form of downloads and a little howto run the script.
Please note that you need to be computer savvy
Regards,
Joop