How Do I Extract Text from Passports?

Passport Structure and OCR

Modern passports follow ICAO Document 9303 standards, which define the layout and data fields for machine-readable travel documents.

The Data Page Every passport has a data page containing: - **Personal details:** Full name, nationality, date of birth, sex - **Document details:** Passport number, issue date, expiry date, issuing authority - **Photograph:** Biometric photo of the holder - **Machine Readable Zone (MRZ):** Two or three lines of encoded text at the bottom

The Machine Readable Zone (MRZ) The MRZ is a standardized text block at the bottom of the data page. It uses a fixed-width font (OCR-B) and follows a strict format:

For a passport (Type P), the MRZ has two lines of 44 characters each: - Line 1: Document type, country code, name, passport number, checksums - Line 2: Passport number, nationality, birth date, sex, expiry date, checksums

The MRZ is designed specifically for OCR reading, so it achieves much higher accuracy than the visual text fields above it.

Accuracy by Passport Type

| Passport Element | Visual Text OCR | MRZ OCR | |---|---|---| | Name | 95-98% | 98-99.5% | | Passport number | 96-99% | 99%+ | | Nationality | 95-98% | 99%+ | | Date of birth | 97-99% | 99%+ | | Expiry date | 97-99% | 99%+ | | Issue date | 95-98% | N/A | | Place of birth | 93-97% | N/A |

Tips for Passport OCR

1. **Photograph the data page flat** — open the passport fully and press it against a flat surface 2. **Avoid glare on the laminate** — the data page has a protective laminate that reflects light. Use indirect lighting or slightly angle the light source. 3. **Capture the entire page** — include all four corners so the engine can detect the MRZ boundaries 4. **Use even lighting** — avoid shadows from the passport cover or your hands 5. **300 DPI minimum** — passport fonts are small and need adequate resolution 6. **Select the correct language** — passports use the issuing country's language plus English/French

Security and Privacy Considerations

Passport data is highly sensitive personal information. When using OCR on passports:

• **Process locally when possible** — FastOCR does not store uploaded documents permanently
• **Delete extracted data** after use — don't keep passport text on shared devices
• **Use for legitimate purposes only** — passport OCR should be used for personal record-keeping, visa applications, or hotel check-ins — not for identity fraud
• **Verify extracted data** — always cross-check OCR output against the original passport for critical applications