Passport OCR

Passport OCR is a document-processing service for extracting structured information from passport scans and assisting with passport translation workflows.

The service uses a multimodal Qwen vision-language model served through Alibaba Cloud Model Studio (DashScope) to recognize passport information. Extracted data is then normalized and validated against document-specific rules, including ICAO 9303 Machine Readable Zone (MRZ) checks and Chinese passport book-number (TIN) validation.

Access to the service is restricted to authorized users.

Features

• Passport field extraction
  Extracts structured information such as passport number, names, nationality, dates, places of birth and issue, issuing authority, book number, and MRZ data.

• MRZ validation
  Validates ICAO 9303 check digits, MRZ structure, field consistency, and common OCR confusions such as 0 and O.

• Book-number validation
  Normalizes and validates Chinese passport book numbers using their expected format and check-digit rules.

• Cross-field consistency checks
  Compares recognized fields against MRZ data and document-specific rules, including names, sex, dates, issuing place, issuing authority, and passport validity period.

• Document orientation handling
  Normalizes document orientation before recognition to improve extraction reliability.

• Structured output
  Recognition results are returned as a validated structured document rather than unstructured OCR text.

• Interactive correction
  The web interface allows recognized fields to be reviewed and corrected manually, with validation updated without requiring another model request.

• Translation workflow
  Extracted passport information can be used to assist in generating standardized Russian-language passport translation documents.

Supported Information

Passport OCR currently extracts and validates information including:

• Passport type
• Country code
• Passport number
• Surname and given names
• Sex
• Nationality
• Date of birth
• Date of issue
• Date of expiry
• Place of birth
• Place of issue
• Issuing authority
• Passport book number
• Holder signature status
• MRZ line 1
• MRZ line 2

Supported Documents

The service is designed primarily around passports within its currently implemented document rules and validation datasets.

Supported input formats include:

• PNG
• JPEG
• WebP
• PDF

Recognition and validation behavior may vary with image quality, document layout, and passport type.

Accuracy and Verification

Passport OCR combines multimodal recognition with deterministic validation rules, but automated extraction is not guaranteed to be error-free.

Users should verify all recognized and translated information against the original passport before relying on the output for official, legal, administrative, or other consequential purposes.

Access

Passport OCR is not an unrestricted public service.

Access is provided only to authorized users. Authentication is handled through Google sign-in, while access authorization is managed separately by the service’s access-control layer.

Data Processing

Passport images and extracted information are processed solely for providing the document-recognition and related translation functionality of the service.

Some processing is performed using third-party infrastructure, including Alibaba Cloud Model Studio (DashScope) for multimodal model inference.

For details about data handling, retention, and third-party processing, see the Privacy Policy.

Policies