Skip to main content

ANPR (License Plate Recognition)

The ANPR action detects and reads license plates from vehicle images, extracting the plate text for logging, access control, or investigative purposes.

Overview​

Automatic Number Plate Recognition (ANPR) is a specialized image analysis technique that:

  1. Locates license plates within an image
  2. Extracts the characters (letters and numbers)
  3. Converts them to searchable text

This action is optimized for Saudi Arabian license plates, which feature both Arabic and English characters in a specific format.

What It Does​

1. Receives Input Image​

ANPR requires images from a previous detection action. It works best with:

  • Images from Object Detection configured to detect vehicles
  • Cropped vehicle images that show the plate area clearly

2. Plate Detection​

The AI locates the license plate within the image by identifying:

  • The rectangular plate shape
  • The characteristic color patterns (white/green backgrounds)
  • The typical plate positioning on vehicles

3. Character Recognition​

Once the plate is located, Optical Character Recognition (OCR) extracts:

  • Arabic numerals and letters (right side of Saudi plates)
  • English numerals and letters (left side of Saudi plates)
  • The complete plate number in a standardized format

4. Data Storage​

The recognized plate text is stored:

  • In the zone_dwell_events table's extra_info field
  • Associated with the detection event for correlation
  • Available for searching and reporting

Configuration​

ParameterTypeRequiredDescription
Input SourcePrevious ActionYesMust follow a detection action that provides vehicle images

ANPR has minimal direct configuration—it automatically processes images from the previous action in the workflow.

Workflow Requirements​

ANPR must be placed after an action that provides images. A typical workflow:

  1. Object Detection: Configure to detect "Vehicle" or "Car"
  2. ANPR: Process the detected vehicle images
  3. Data Storage/Action: Log the plate or take action based on the read

Understanding Results​

ResultMeaningAdditional Info
processing_completePlates were found and readPlate text stored in event record
no_plate_detectedNo license plate found in imageMay indicate poor image quality or no visible plate
read_failedPlate detected but characters couldn't be readUsually due to blur, angle, or obstruction

Common Use Cases​

Parking Access Control​

  • Setup: Camera at parking entrance, detect vehicles, read plates
  • Integration: Compare against authorized plate list
  • Action: Grant or deny access based on plate recognition

Vehicle Logging​

  • Setup: Camera at facility entrance/exit
  • Schedule: Continuous during operational hours
  • Use: Maintain log of all vehicles entering/exiting for security

Investigation Support​

  • Setup: Wide-area camera monitoring
  • Trigger: Triggered by security events or manual request
  • Use: Capture plates of vehicles present during incidents

Visitor Management​

  • Setup: Camera at visitor parking
  • Integration: Pre-registered visitor plates
  • Use: Verify expected visitors and flag unexpected vehicles

Troubleshooting​

No Plate Detected​

  1. Camera Positioning: ANPR cameras should be positioned specifically for plate capture:

    • Low angle (near plate height, typically 0.5-1 meter)
    • Direct view of plate (not extreme angles)
    • Appropriate distance (plates should fill significant frame area)

  2. Image Quality: Plates must be clearly visible:

    • Minimum 100 pixels plate width recommended
    • Sharp focus on the plate area
    • Adequate contrast

  3. Lighting: License plates are often reflective:

    • Avoid direct sunlight causing glare
    • Use infrared illumination for night operation
    • Consider anti-glare camera positioning

  4. Vehicle Speed: Fast-moving vehicles cause motion blur:

    • Use cameras with fast shutter speeds
    • Position cameras where vehicles naturally slow down (gates, speed bumps)

Characters Not Read Correctly​

  1. Dirty Plates: Encourage clean plates or accept partial reads for dirty conditions.

  2. Plate Damage: Bent, faded, or damaged plates reduce accuracy.

  3. Non-Standard Plates: Decorative frames or covers that obscure characters will cause read failures.

  4. Foreign Plates: This action is optimized for Saudi plates. Plates from other countries may have lower accuracy.

  5. Character Confusion: Similar characters (0/O, 1/I, 8/B) can be misread. Review results for patterns.

Slow Processing​

  1. Image Size: Very high-resolution images take longer to process. Consider capturing at optimal resolution for ANPR (720p-1080p).

  2. Multiple Plates: If multiple vehicles/plates are in frame, processing takes longer.

  3. Server Load: Ensure adequate server resources for the AI processing.

Inconsistent Results​

  1. Lighting Changes: Results may vary between day and night. Consider separate day/night configurations.

  2. Weather Conditions: Rain, fog, and dust affect visibility. Accept lower accuracy during adverse conditions.

  3. Camera Drift: Verify camera position hasn't shifted, especially after maintenance.

Best Practices​

Camera Setup​

  • Dedicated ANPR Camera: Use a camera specifically positioned and configured for plate reading, separate from general surveillance cameras.

  • Infrared Capability: For 24/7 operation, use cameras with IR illumination that can read plates in darkness.

  • Optimal Angle: 20-30 degrees from horizontal, capturing plates head-on as vehicles approach.

  • Appropriate Zoom: Plates should occupy 10-15% of frame width for optimal reading.

Operational Considerations​

  • Speed Control: Position cameras at natural slowdown points (gates, barriers, speed bumps).

  • Single Lane: Capture one lane at a time to avoid plate overlap.

  • Weather Protection: Protect cameras from rain and condensation.

Data Management​

  • Logging Policy: Determine how long plate data should be retained based on your requirements.

  • Access Control: Limit access to plate data to authorized personnel only.

  • Integration: Consider integrating with existing access control or parking management systems.

Technical Notes​

  • Supported Plates: Optimized for Saudi Arabian plates (both private and commercial formats).

  • Character Set: Recognizes Arabic numerals (٠-٩), Arabic letters, and English alphanumerics.

  • Processing Time: Typical processing is under 2 seconds per image on standard hardware.

  • Accuracy: Under optimal conditions, expect 95%+ accuracy on Saudi plates.