AI Guardrails:Image OCR

Prerequisites

Before you begin, make sure that you have:

• Activated the AI Guardrails service in the Alibaba Cloud console. If the service is not activated, API calls return error code 596

• The AI Guardrails SDK for PHP and its dependencies installed. For details, see Installation

• The PHP version specified in the Installation topic. Using an unsupported PHP version causes operation calls to fail

• An AccessKey pair (AccessKey ID and AccessKey secret) for a Resource Access Management (RAM) user, stored in environment variables

Input requirements

The SDK accepts only image URLs. Local files and binary data are not supported.

Submit a synchronous OCR task

The ImageSyncScanRequest operation submits a synchronous OCR task. Set the scenes parameter to ocr to recognize text in images.

Supported regions: cn-shanghai, cn-beijing, cn-shenzhen, ap-southeast-1

The following example recognizes the front of an ID card. To detect other card or certificate types, change the card value in the extras parameter.

<?php

use AlibabaCloud\Client\AlibabaCloud;
use AlibabaCloud\Green\Green;

try {
    // Reuse this client instance across requests to avoid repeated connections.
    AlibabaCloud::accessKeyClient(
            getenv("ALIBABA_CLOUD_ACCESS_KEY_ID"),
            getenv("ALIBABA_CLOUD_ACCESS_KEY_SECRET")
        )
        ->timeout(10)          // Request timeout: 10 seconds (client default)
        ->connectTimeout(3)    // Connection timeout: 3 seconds (client default)
        ->regionId('cn-shanghai')
        ->asDefaultClient();

    // Each task requires a dataId (business data identifier) and a url (image URL).
    $task1 = array(
        'dataId' => '<your-data-id>',
        'url'    => '<your-image-url>'
    );

    // Specify the card or certificate type to recognize.
    // Example: front of an ID card
    $extras = array('card' => 'id-card-front');

    /**
     * One image = one task. To detect multiple images, create multiple tasks.
     *
     * Batch detection takes longer on average than single-image detection.
     * The more images in a batch, the higher the probability of increased
     * response time.
     */
    $result = Green::v20180509()->imageSyncScan()
        ->timeout(10)          // Request timeout: 10 seconds (overrides client default)
        ->connectTimeout(3)    // Connection timeout: 3 seconds (overrides client default)
        ->body(json_encode(array(
            'tasks'  => array($task1),
            'scenes' => array('ocr'),
            'extras' => $extras
        )))
        ->request();

    print_r($result->toArray());

} catch (Exception $exception) {
    echo $exception->getMessage() . PHP_EOL;
}

Replace the following placeholders with your values:

Billing

OCR detection is billed based on the number of detected images multiplied by the unit price for the detected card or certificate type. For pricing details, see the AI Guardrails pricing page.

Performance tips

• Reuse the client instance. Creating a new client per request adds connection overhead. Initialize the client once and reuse it across calls.

• Prefer single-image detection for latency-sensitive workloads. Batch detection averages a longer response time than single-image detection, and response time increases with batch size.

References

• Image OCR detection API

• PHP SDK installation