Driver's license OCR: fields, the PDF417 barcode and how to automate license capture

Driver's license OCR reads the printed front of a driver's license and turns it into structured data: name, date of birth, address, license number, class, and the issue and expiration dates. On US licenses it's paired with a decoder for the PDF417 barcode on the back, which stores the same core fields under AAMVA element IDs, so the two copies can be checked against each other. Banks, lenders, insurers and trucking fleets use it to fill onboarding, KYC, loan and driver records without retyping the card.

This guide covers the fields, the barcode, how extraction works, the checks to run and the tools.

What is driver's license OCR?#

OCR turns the characters in an image into text. Driver's license OCR adds the part that makes the text useful: it knows which string is the license number and which date is the date of birth, and returns each one as a named field. It's also called driving license OCR, or sold as a driver's license data extractor or parser.

A US license carries its data twice. The front is printed for people. The back has a PDF417 barcode for machines, which a barcode reader decodes, not OCR. Holograms and microprint are security features, not data to extract.

Fields you can extract from a driver's license#

The AAMVA DL/ID Card Design Standard, the North American standard for license design, sets the fields a license must show. It numbers them on the card (4d for the license number, 3 for the date of birth) and gives each a three-letter element ID in the barcode.

• Name

  Family name (field 1) and given names (2). Barcode: DCS, DAC and DAD, plus flags that show whether a name was cut short.

• Date of birth

  Field 3, written MM/DD/CCYY on US cards. Barcode: DBB.

• Address

  Field 8: street, city, state and ZIP code. Barcode: DAG, DAI, DAJ and DAK.

• License number and issuing state

  Field 4d, an ID each state assigns in its own format, and the state named across the top. Barcode: DAQ, with the issuer identified in the header.

• Issue and expiration dates

  Fields 4a and 4b. Barcode: DBD and DBA.

• Class, endorsements and restrictions

  Fields 9, 9a and 12, in the state's own codes, explained on the card. Barcode: DCA, DCD and DCB.

• Physical description

  Sex (15), height (16) and eye color (18). Barcode: DBC, DAU and DAY.

• Document discriminator

  Field 5, a number that tells this card apart from others issued to the same person. Barcode: DCF.

These fields also cover what a bank's Customer Identification Program (CIP) must record about an ID it relies on: document type, ID number, place of issuance, and issue and expiration dates.

What's in the barcode on a driver's license?#

The barcode on the back is a PDF417, a two-dimensional barcode that every AAMVA-compliant and REAL ID-compliant card must carry. Its standard fields aren't encrypted, so any PDF417 reader can decode them: a header that starts with "@" and "ANSI" and identifies the issuer by a six-digit number, then one field after another, each opening with its element ID.

The barcode repeats the front, but not always in the same form, so normalize before you compare:

Canadian cards write barcode dates year first (CCYYMMDD). A mandatory field with no value reads "NONE", or "unavl" when the issuer doesn't have the data.

How to extract data from a driver's license with OCR#

Whatever tool you use, license capture runs through the same stages, from upload to your KYC or loan system.

1. Capture both sidesFlat, in focus and without flash: glare on the laminate can wash out a field or the barcode. Phone photos work when the whole card is in frame.
2. Identify the state and versionLayouts differ by state, and older designs stay valid after a redesign: AAMVA's standard cites estimates of well over 200 valid designs. The barcode header names the issuer and the standard version.
3. Read the frontOCR reads the text, and extraction maps each value to its field by the label or reference number beside it.
4. Decode the barcodeA PDF417 reader turns the back into text, and a parser splits it into fields on the element IDs.
5. Cross-checkNormalize formats, then compare the front with the barcode and both with the application.
6. Review and deliverUnreadable or mismatched fields go to a person with the image beside them; clean records go to your KYC system or loan origination system (LOS) through an API.

Checks to run on every driver's license#

These are data checks: they catch misreads and mismatches between the license and what the applicant told you. Proving the card is genuine takes an identity verification tool, and confirming the state holds the same data takes a check such as AAMVA's DLDV service.

• Not expiredThe expiration date is after the day you rely on it. The CIP rule names an unexpired government-issued photo ID, such as a driver's license, as a document banks may use.
• Front matches barcodeLicense number, name, date of birth and expiration agree once formats are normalized. A mismatch usually means a misread, so re-read the field or send it for review.
• Matches the applicationName and date of birth agree with the application and the rest of the file. A different address may only mean the person moved after the license was issued, so ask rather than decline.
• License number fits the stateEach state sets its own format, so check the length and pattern for the issuing state.
• Nothing missingBoth sides arrived, the barcode decoded and every field your records need was read.

Here's the front-against-barcode check on one license:

Driver's license OCR APIs, SDKs and tools#

Three kinds of tools read driver's licenses, through an API, an SDK or both. Which fits depends on whether you need the whole file read, the license alone, or proof that the person matches the ID.

• Document AI platforms

  Docsumo, our product, reads the license with the rest of the file, such as the application, pay stubs and bank statements, and on the Enterprise plan checks them against each other. Output goes through an API and webhooks. It doesn't match selfies or test a card's security features.

• Cloud OCR APIs

  Amazon Textract (AnalyzeID), Google Document AI (US driver license parser) and Azure Document Intelligence (ID document model) return the fields as JSON, with published pricing. You build the checks and workflow around them.

• Identity verification platforms

  Jumio, Socure and Stripe Identity check the ID document and match a selfie to its photo while a customer signs up, and return a verification result.

For a vendor-by-vendor comparison of the identity tools, see the best KYC document verification software.

How to automate driver's license capture#

A license rarely arrives alone. Docsumo reads the front with a pre-trained driver's license model, and a step in the workflow decodes the PDF417 barcode on the back with a barcode library, so the two can be compared. It handles any other document type, so it reads the license with the rest of the loan file or onboarding packet. On the Enterprise plan, cross-document validation checks the license against the other documents: the same name and date of birth everywhere, and the address against the application.

Fields below the confidence threshold you set go to your reviewers, who click a field to see its source line on the image; their corrections improve the model. Results reach your KYC system or LOS through the API and webhooks. For motor carriers, driver qualification file processing flags an expiring CDL against a window you set when the record is read.

• 99%99%field-level accuracy across 250+ document types
• 95%+95%+of documents processed straight through, without manual review

The bottom line#

A US driver's license carries its data twice: printed on the front and encoded in the PDF417 barcode on the back. Read both, normalize the formats, check them against each other and the application, and send only the mismatches to a person. When you also need proof that the card is genuine, add identity verification.

Book a demo with a sample onboarding or loan file, or start a free trial.

Frequently asked questions#