Text recognition and field mapping

Vehicle registration OCR: from recognised text to finished fields

Upload a photo, a scan or a PDF of the German Zulassungsbescheinigung Teil I and get finished, named fields back instead of loose text: licence plate, date of first registration, manufacturer key, type key, VIN, make, model and more — up to 65 details per document. For every value you also see how sure the recognition was and where on the document it stands. Fahrzeugschein24 is built for this one form, older editions included.

Start for freeRead the API documentation

OCR returns words — you need fields

Plain OCR answers one question: which characters appear where in the picture? On the Zulassungsbescheinigung Teil I the meaning sits in the box a value is printed in — four digits on their own could be a manufacturer key, an engine capacity or a towing weight, and the labels are short codes rather than words.

That mapping is what Fahrzeugschein24 does for you — also when the document is photographed at an angle, folded across a line, or catching a reflection off the security print. It saves you the laborious half of the job: the step from recognised text to finished fields. Reading a Fahrzeugschein shows what that looks like day to day.

Six places where the work is already done for you

On the documentPlain text recognitionFahrzeugschein24
Manufacturer key, four digits in a boxThe four digits as loose textThe manufacturer key, named as such
Type key, printed with further code lettersOne block of characters as printedThe type key on its own, variant and version separated
Licence plateThe line with the form's separatorsThe licence plate in one consistent spelling, every time
Date of first registrationA date as text, in the form's formatFirst registration as a date you can pass into your system
VIN and the printed check digitTwo pieces of text in two placesThe VIN, checked against the printed check digit
Energy source and fuel in plain wordsTwo values whose relationship stays openBoth entries named, with a note if they disagree

One photo in, finished vehicle data out

Document in:   photo of a Zulassungsbescheinigung Teil I

Details out (extract):

  Licence plate                M SN 1234 E
  VIN                          WVWZZZ1KZAW123456
  Check digit matches          yes
  Manufacturer key             0603
  Type key                     BFT
  First registration           14.03.2019
  Fuel / energy source         BENZIN
  Keeper, address, dimensions,
  weights, axles, tyres …      as printed on the document

You only check what stands out

Every single value comes with an indication of how sure the recognition was. You set the confidence level at which a value goes to a person for a quick look, instead of re-reading every result in full.

The location comes with it: you see where on the document a value stands, and can switch on the matching image detail in your account. A doubtful value is settled in seconds. And a poorly lit photo usually affects individual entries, not the whole result — the confidence value shows which ones.

Checks that run for you on every document

The VIN against the printed check digit

The vehicle identification number is compared with the check digit printed on the same document. If the two do not fit together, you are told — typical for transposed digits. Reading the VIN explains how that check digit works.

Plate, dates and key numbers in one spelling

The licence plate, dates and key numbers come back in a fixed spelling, the plate as "M SN 1234 E". The district code is matched against the German district codes, so mix-ups between similar letters stand out.

Contradictions are reported, not corrected

When two entries do not fit together — the energy source given once as a code and once in words, say — you get a note. Both values stay as they were read; the call stays with you.

Image anomalies from Business up

From the Business plan on, a check for image manipulation runs alongside and reports anomalies in the uploaded picture. It is a pointer for your own review, not an official determination of authenticity.

Built for the German document

Fahrzeugschein24 is built precisely around the German Zulassungsbescheinigung Teil I and the older Fahrzeugschein — their boxes, their short code labels, their different issue years. Accepted formats are JPEG, PNG, HEIC, PDF, TIFF, BMP and WebP up to 20 MB; with a PDF the first page is read. A phone photo is normally enough — the scanner page shows what matters when taking it, and the REST interface wires the whole thing into your own process.

Processing runs on servers in Germany and therefore inside the EU. Uploaded documents are not stored; what remains is the extracted data and the account records. The Article 28 GDPR data processing agreement is available without asking, and further detail is on security and data protection.

Frequently asked questions

Why is plain text recognition not enough for a German registration document?

Because it gives you words, not named entries. On this form the meaning sits in the box a value is printed in: four digits on their own could be a manufacturer key, an engine capacity or a towing weight. Here you get that mapping and the consistent spelling already done.

How many details do I get per document?

Up to 65. How many it actually is depends on what is printed on the document: anything not printed there comes back empty rather than filled with a guess.

How do I know whether I can trust a value?

Every entry comes with an indication of how sure the recognition was and where on the document it stands; the image detail can be switched on in your account. You set the confidence level at which a person takes a look — and that look goes to one snippet, not the whole document.

What does the check-digit comparison tell me?

Whether the VIN that was read matches the check digit printed on the same document. If it does not, you get a note — typical for transposed digits. If the document carries no usable check digit, the result says so plainly.

Can I try it on my own document first?

Yes. The free account gives you 15 extractions a month, no credit card. What comes after that is on the pricing page, and more answers are in the FAQ.

Compare it on a document you know

The difference between raw text and finished fields shows fastest on a registration document whose contents you already know. The Free plan costs 0 EUR, includes 15 extractions per month and needs no credit card.

Create a free account