The FLOW framework enhances traffic monitoring by integrating the advanced Automated License Plate Recognition (ALPR) technology, enabling seamless vehicle identification and tracking. This technology links detected vehicle trajectories with their actual license plates, which offers a wide range of application scenarios.

The FLOW framework utilizes ALPR for vehicle monitoring, speed measurement, and parking management, while also enhancing the detection of various traffic events—such as wrong-way driving, red-light running, U-turns, and illegal lane changes—by linking them to specific vehicles through their license plates.

Usually, multiple cameras are installed on-site to optimize coverage for different traffic views, as well as to cover the ALPR function itself. Together, these multi-camera setups provide a comprehensive view of the traffic situation at a specific location. Moreover, ALPR plays an important role in analyzing vehicle movement patterns and travel times between different locations.

Visual evidence collection further extends ALPR capabilities by providing documented proof of detected traffic events. When combined with ALPR, the Evidence feature delivers a clear record of specific vehicles’ movements.

High-quality cameras significantly enhance recognition accuracy, especially under low-light conditions or challenging viewing angles. The system detects and reads license plates only on identified traffic objects. If the primary object detector fails to recognize a vehicle, its license plate will not be processed. Proper camera placement and configuration are therefore essential for optimal results. Further recommendations on camera positioning, viewing angles, and setup are provided below.

Figure 1: A suitable snapshot (OK) for license plate reading under poor lighting conditions – vehicle contours are still visible, license plates have clearly defined edges, and the characters on them are readable.

Figure 2: An unsuitable snapshot (Not OK) for license plate reading – vehicle contours are no longer apparent. The vehicle may not be detected, and its license plate might not be processed.

In FLOW, the ALPR engine is activated individually for each video analytics through the client application or API. Enabling ALPR typically increases processing demands by 15-30%, depending on the number of vehicles tracked. The system efficiently balances performance and ensures real-time results by optimizing the processing of detected objects. Operators can configure ALPR settings to prioritize specific detection zones, reducing unnecessary processing in low-traffic areas.

Figure: Example of activating the ALPR engine through FLOW Insights.

To comply with privacy regulations, FLOW offers on-the-fly license plate hashing, ensuring that license plate data remains anonymized and cannot be retrieved in its original form. The hashed representation is unique enough for tracking purposes while maintaining compliance with GDPR and similar privacy regulations. FLOW also allows users to customize data retention policies based on local regulations and operational needs.

Figure: Example output showing active Hashing in FLOW.

License plate data can be filtered using Matching and Not Matching modes, allowing operators to filter or flag specific vehicles or vehicle groups, define zone entry access permissions, and similar restrictions.

Figure: Example of how to integrate the operator for license plate–based filtering, including the configuration dialog.

For effective license plate recognition, follow the below guidelines to achieve the best results:

Further setup recommendations and tips may vary based on the camera manufacturer specifications, the specific installation location, and the unique analysis requirements.

Below, we provide a visual guide showing the relationship between the camera’s installation height, its view angle settings and the readability of license plates.

FullHD ALPR, 40° HFOV (zoomed-in)

4K ALPR, 40° HFOV (zoomed-in)

Figure: Relationship between the camera’s installation height, its view angle relative to the vehicle's travel direction, and the readability of license plates in pixels.

Following the overview above, for a FullHD setup (first table) with the camera installed at a height of 6 meters, license plates are readable (assuming 16 pixels or more) for vehicles captured at angles between 29-37 degrees and within a horizontal distance of 8-11 meters from the camera. The camera should be framed accordingly.

FLOW supports license plate recognition across all EU countries and most alphanumeric-only plates worldwide, reading them from left to right and top to bottom. The system is also adaptable to multi-line plates, ensuring reliable detection across different plate formats.

Figure: Examples of license plates recognizable by the FLOW ALPR Engine.

Additionally, the system can be customized on demand to recognize non-alphanumeric plates and specific vehicle markings, such as hazardous material labels or fleet identification codes, for specialized use cases.

The ALPR engine in FLOW is a powerful feature that helps traffic operators maximize their insights by providing high-accuracy license plate recognition in real-time. It enhances traffic monitoring and security applications, allowing for seamless multi-camera data correlation and integration with enforcement systems through documented REST API or webhooks. The system can be integrated with various platforms while ensuring compliance with GDPR through its anonymization feature. Follow proper camera setup and positioning guidelines to achieve the best results.

We will be happy to answer any questions you may have. Contact us!