ALPR / ANPR Definition
ALPR and ANPR are interchangeable acronyms used to describe automatic license plate recognition systems. ALPR stands for Automatic License Plate Recognition, while ANPR stands for Automatic Number Plate Recognition. In practice, both refer to a set of image-processing methods that detect a license plate in a photo or video recording, locate its position, and then read the characters using OCR or machine learning models.
In the context of photo and video anonymization, ALPR / ANPR has a dual meaning. First, it refers to the technique used to identify license plate numbers. Second, it describes a category of processing that may involve personal data if a plate number makes it possible to identify a natural person directly or indirectly. This interpretation follows Article 4(1) GDPR and the broad understanding of identifiability adopted by data protection authorities and EU case law.
From a technical perspective, an ALPR / ANPR system usually works in several steps: license plate detection, image normalization, character segmentation or sequential recognition, syntax validation of the plate number, and result storage. In materials intended for publication or external sharing, the same detection stage may be used not to read the plate, but to blur or mask it. This distinction is important from a compliance perspective. Reading the plate expands the scope of processing. Mere detection followed by immediate redaction can support the data minimization principle under Article 5(1)(c) GDPR.
When ALPR / ANPR Becomes Personal Data Processing
Not every vehicle registration number will always qualify as personal data in every situation. The assessment depends on the context, the purpose, and the ability to link the plate number to a specific person. Across the EU, a precautionary approach generally prevails: if the controller, or another entity with which the data may be combined, has a realistic ability to identify the owner or user of the vehicle, the registration plate should be treated as information relating to an identifiable person.
In Poland, there is some divergence in interpretation. On the one hand, the Polish data protection authority (UODO), the EDPB, and practices based on a broad understanding of identifiability support treating license plates as personal data in certain contexts. On the other hand, some administrative court rulings have held that a registration number alone does not always constitute personal data. For photo and video materials intended for publication, the safer approach is to anonymize license plates, especially where the vehicle image is linked to a location, time, route, or other identifying features.
The CJEU judgment in case C-582/14 Breyer is also relevant, as the Court adopted a broad approach to identifiability using additional information available to other parties. The case did not concern license plates specifically, but it remains important when assessing indirect identification. In compliance practice, a license plate visible in video footage should be analyzed in a similar way to other indirect identifiers.
How ALPR / ANPR Works in Photo and Video Processing
In modern systems, license plate detection is usually based on deep learning models. This is also important for anonymization. To automatically blur license plates, a pre-trained AI model is needed—one that has learned to recognize the location of plates across varied images. Only such a model can then be used operationally to detect and cover license plates in video files or photos.
A typical technical pipeline includes the following stages:
- object detection — identifying the bounding box around the plate,
- tracking across frames — maintaining mask continuity in video footage,
- optional OCR — reading the characters if the goal is plate recognition,
- anonymization — blurring, pixelation, or full masking of the detected area,
- quality control — verifying missed detections and false positives.
In Gallio PRO, automation applies only to faces and license plates in stored files, not to real-time anonymization or live video streams. The software does not automatically detect logos, tattoos, name badges, documents, or content displayed on monitors. Such elements can be blurred manually in the editor.
Key ALPR / ANPR Parameters and Metrics
Evaluating an ALPR / ANPR system requires separating detection performance from recognition performance. In compliance-focused use cases, perfect OCR is often less important than high detection recall, because an undetected plate remains unanonymized.
The table below summarizes the core metrics.
Parameter | Meaning | Practical relevance for anonymization
|
|---|---|---|
Detection recall | The percentage of license plates correctly detected | Critical — low recall increases the risk of data disclosure |
Detection precision | The percentage of detections that are actual license plates | Affects the number of unnecessary masks |
IoU | The degree of overlap between the detection box and the object | Affects whether the mask covers the entire plate |
OCR accuracy | The correctness of character recognition | Secondary if the goal is only to blur the plate |
Latency per frame / file | The time required to process the material | Affects the efficiency of the offline workflow |
False negative rate | The percentage of missed license plates | A key indicator of legal risk |
Basic performance can be described using the following formulas:
Precision = TP / (TP + FP)
Recall = TP / (TP + FN)
Where TP means true positives, FP means false positives, and FN means missed license plates. In anonymization workflows, higher recall is usually preferred even at the cost of a moderate increase in false positives, because excessive blurring is generally less risky than failing to blur a plate at all.
Legal Obligations and GDPR Compliance
If ALPR / ANPR is used to read license plates and further use the result, the controller must determine the legal basis for processing, the purpose, the retention period, and the categories of data recipients. Information obligations must also be fulfilled unless a GDPR exception applies, and the risk to individuals’ rights and freedoms must be assessed. In the case of monitoring, traffic analysis, or debt enforcement, a Data Protection Impact Assessment (DPIA) under Article 35 GDPR will often be required.
For publishing photos and video recordings, compliance obligations often lead to a different conclusion: instead of reading the plate number, it should be blurred without delay. This approach limits the scope of data and supports privacy by design under Article 25 GDPR. In many European countries, blurring license plates in published materials is widely accepted as a practical standard based on national rules, regulatory guidance, and the EU approach to privacy protection.
ALPR / ANPR Challenges and Limitations
License plate recognition performance depends heavily on the quality of the source material. Problems arise with strong compression, motion blur, high contrast, dirty plates, unusual character fonts, and partial occlusion. For anonymization, challenges also include small background objects, nighttime footage, and light reflections.
In practice, the following limitations should be taken into account:
- a model trained on one country may perform worse on other license plate formats,
- low resolution reduces both detection performance and OCR accuracy,
- false negatives are the most serious error from a data protection perspective,
- OCR itself is not necessary for anonymization and may unnecessarily expand the scope of processing.
Normative References and Sources
The term ALPR / ANPR is not defined in a single EU legal act as a formal compliance concept. Its assessment is based on general rules governing personal data, privacy, and the secure processing of visual material.
- Regulation (EU) 2016/679 — GDPR, in particular Articles 4, 5, 25, 32, and 35.
- Court of Justice of the European Union, judgment of 19 October 2016, C-582/14, Breyer — broad interpretation of identifiability using additional information.
- European Data Protection Board (EDPB) — guidance on the concepts of controller, processing, and risk assessment, used as supporting references when analyzing identification systems.
- ISO/IEC 27001:2022 — requirements for an information security management system, relevant for organizations processing photos and video.
- ISO/IEC 23894:2023 — guidance on risk management in AI systems, useful when assessing license plate detection models.