Both TrafficSurvey and FLOW have license plate reading functionality and their LPR capabilities have a couple of things in common. One of the commonalities is that once a license plate has been read it is associated with the trajectory for its whole duration meaning you can tell what is the license plate of a car even though its LP is no longer visible in the camera view. License plate reading must be done from light views from cameras ideally in 9-15 meter height so that the camera can see the license plate well. Both FLOW and TrafficSurvey engines are quite flexible in terms of the angle at which it is able to read the license plate. As you can see in the GIF below

FLOW

FLOW uses a different engine from TrafficSurvey. FLOW for TrafficEmbedded and TrafficEnterprise have a built-in LPR that is able to recognize all European and some Europe-similar license plates. We are flexible in terms of product development and we offer to train our neural networks and add license plate reading support for a given country based on your request. It would also be possible for example to read numbers on trains or shipping containers if you would like this feature implemented. FLOW also has advanced functions that allow the creation of license plate whitelists and blacklists for example for finding lost vehicles or limiting access to a parking lot for residents only.

TrafficSurvey

Which countries are supported?

TrafficSurvey can detect all EU license plates as well as most alphanumerical-only license plates. See the image below for examples of license plates that the TrafficSurvey LPR service can read. Please note that in order to read a license plate reliably, it should be at least 18pixels in the vertical dimension but this is easily achieved for most videos. You can also test whether license plates will be read in your video for FREE by ticking the LPR option when doing the 5-minute free sample analysis. Please note that there a limit of 10x free sample analyses per 24 hours.

To offer higher privacy protection in regard to GDPR TrafficSurvey also allows irreversible License Plate Number hashing.

How to make a suitable video for License Plate recognition?

Here are general guidelines applicable for both FLOW and TrafficSurvey to ensure high reliability of the number plate detection:

Good enough resolution specifically the height of the letters and signs on the license plates should be at least 18 pixels in order to read the license plate reliably
Use High-quality video output (FullHD or 4K)
Number plates must not be blurred and must be easily readable and visible to the naked eye
Number plates must not be overexposed (with too much light) or underexposed
Number plates must not be deformed (for example by interleaving or by the rolling shutter effect)

Example picture where LPR is possible. Ideally, the camera should be higher than this.