Traffic Surveillance

Follow the process step by step

The starting point is a picture of the vehicle, often taken by a speed camera. Using advanced image processing and trainable search algorithms, our engine finds areas which contain the sought information with high probability. The result of this step is a sorted list with the positions of license plates or faces.

These images are then processed by PLANET AI‘s Intelligent Text Recognition which uses state-of-the-art machine learning and artificial intelligence technologies. The result is our patented PerceptionMatrix for each position – containing all probabilities for the given items. For the vast majority of countries, there are precise rules according to the structure of a license plate. These are used in addition to the PerceptionMatrix which ensures that correct data is read even under unfavorable conditions.

The faces found are assigned a probability as to whether it is the face of the driver or other passengers. In Germany, only the driver is liable for an offense, while the faces of other passengers must be made unrecognizable for data protection reasons.

Images from traffic surveillance typically have a high dynamic range – often very bright license plate compared to the dark interior, including the passengers. In this step, both the overall image and individual images of specific areas (license plate, faces) are adjusted to display the details of interest as well as possible.

Due to privacy regulations, we have to disguise several faces and license plates.