ZIRflow — czyli analityka wideo, a nie wideodetekcja

AI searches the image for characteristics of objects learned during DNN training. A change in the image, for example, caused by the shadow of a car, does not activate the zone. Only the identification of an object in the zone, which the neural network knows about, activates it. Each object is then classified and tracked, and its movement is recorded as a trajectory. If the object appears in a user-designated detection zone, information about it is sent to the traffic light controller.

ZIRflow detect objects into 17 categories (passenger car, van, small truck, motorcycle, bicycle, pedestrian, wheelchair, animal, etc.) The system collects information about the characteristics of detected objects, such as colour, the trajectory of passage, speed or acceleration. Traffic trajectory analysis can be used to make the distribution of traffic streams at an intersection.

Video analytics system a step forward

AI can show real-time images from multiple cameras. It 'captures' all objects moving between the fields of observation of the various image recording devices. This solution enables observation of traffic flow, for example, on an entire roundabout from several cameras installed centrally on a mast in the middle of the roundabout. The image from the camera makes it possible to analyse the distribution of traffic flows by vehicle class, colour, speed and other features read by AI.

In addition to checking for the presence of vehicles in the junction area and the analysis of the traffic flow, the classification of detected objects enables giving selective priority to a not-moving tram on a dedicated track. If ZIRflow detects a tram in the traffic flow, it transmits the proper information to the controller and then gives the priority.

Costs

ZIRflow allows to communicate with external devices using several different communication protocols based on Ethernet. For example, we use a UDP-based protocol to communicate with the traffic light controller. As a result, the ZIRflow system does not require additional output modules and the controller with input modules. It keeps the cost of installing ZIRflow at the level of the cost of existing video detection systems.

The ZIRflow system is a loT solution and makes it possible to connect IP cameras currently in use by a potential Investor.