Tencent Cloud Multi-Region Passport OCR

Problem

Passport recognition is more structured than general document OCR. In real workflows, teams often need to extract fields such as name, passport number, date of birth, nationality, and validity dates from passports issued by different countries, crop the holder image, and detect risks such as copies, rephotographed images, PS edits, glare, or blur.

How It Works

This skill wraps the Tencent Cloud MLIDPassportOCR passport recognition API. It supports mainland China, China Hong Kong/Macao/Taiwan regions, and other countries or regions, covering 80+ countries/regions with machine-readable passports. Input can be ImageBase64 or ImageUrl; at least one is required, and ImageUrl takes priority when both are provided. Setting RetImage=true requests the holder image as Base64 data. The skill selects the domestic or international request path, parses both the MRZ machine-readable zone and the VRZ visual zone, and returns structured JSON results.

Key outputs include passport number, surname, given name, full name, date of birth, sex, expiration date, issuing country, nationality, and document type. For Thai passports where IssuingCountry=="THA", it additionally validates consistency between OCR fields and MRZ fields. For mainland China passports, Signature, IssuePlace, and IssuingAuthority return meaningful values. When requesting the ap-bangkok region, the CardCount field is also returned.

Scope and Notes

This skill is best suited for identity document extraction, form prefill, and data verification workflows. Be aware that WarnCardInfos alerts only apply to international station requests and return an empty array for domestic requests. The Warn and AdvancedInfo fields are deprecated. Images must meet size, download-time, and resolution constraints, and PDF input uses only the first page. The default rate limit is 5 requests per second, and TENCENTCLOUD_SECRET_ID and TENCENTCLOUD_SECRET_KEY must be configured before use.