Performed by AutoSDK recognition engine, LPR is the main feature of every principal VIT product, whether it's a plug-in for VMS or a hardware device for on-site traffic surveillance. As a core component, AutoSDK determines three fundamental licensing parameters:
number of LPs issuer-states;
number of recognition channels;
video processing speed.
Now, let's consider them in more detail.
When preparing a license for you, we will have to specify for AutoSDK what kind of LPs are to be searched on your video. Those LP templates are grouped by the country where they were issued. So what are those countries and their number make the first licensing parameter.
We call one traffic lane under video surveillance a recognition channel. Also, we sometimes call it a recognition thread, because it usually takes one OS thread to recognize LPs within one traffic lane. Depending on resolution, your cameras may capture one or multiple lanes in a frame, thus, your license will provide the corresponding number of available recognition channels (see the Example 1).
Video processing speed is a number of frames to be processed by AutoSDK every second. In other words, AutoSDK selects a specific number of frames from a frameset provided every second and performs search and recognition operations on those frames. The rest frames are overpassed, which allows for AutoSDK load balancing. This feature helps to utilize one's resources optimally, as, in some installation zones, analysis of every single frame from a camera is not necessary for high LPR quality (for example, when each vehicle stops in front of a camera for a while before leaving the frame). So, AutoSDK has the following operation modes: