On this page (7)
Read accuracy, permission logic and event logging are a matter of configuration and work the same way in both variants.
Edge and central ANPR compared
| Edge | Central | |
|---|---|---|
| What crosses the network | the read result with a frame thumbnail and a crop of the plate | the video stream |
| Link needed to the gate | minimal | stable, with enough bandwidth |
| Latency | lowest, the read is produced at the gate | depends on the network |
| When the link fails | the device keeps working and buffers events | no reads |
| Hardware at the gate | a computing device in the cabinet | the camera only |
| Updates | on every device | in one place |
| Upfront cost | higher by the cost of the devices | lower |
When to choose edge ANPR
Reading plates at the gate pays off above all where the link is weak. A gate at the far end of the site, connected by a radio link or over LTE, will struggle to carry a video stream, while a read result with a frame thumbnail gets through without any trouble. The second reason is operation during a network outage: the edge device carries on recognising plates and buffers the events, which reach the server once the connection is back.
With many gates spread across a site, the site network matters as well, because the combined video stream from ten points can load it more heavily than you would expect. Then there is latency: the difference is a fraction of a second, but at a barrier it can be noticeable.
When central analysis is enough
Analysis on the server is enough when the site has one or two gates close to the server room. The site network will then carry the video with no noticeable load, and maintenance is simpler, because there is one place to update and less hardware out in the yard. You also don't have to buy edge devices, which lowers the upfront cost. Hardware is always priced separately from the licence, and the pricing configurator shows the licence amount.
With one or two gates within reach of a good network, the advantage of the edge approach is therefore theoretical. It starts to count above a handful of gates and wherever the link to the gate is weak.
What the choice of variant does not change
Where the plate is read has no effect on read accuracy. That comes down to how the camera is mounted, the lighting and the condition of the plate. The choice doesn't change the permission logic either: the decision to open the gate is made by rules, regardless of where the number was recognised.
The ANPR variant has no bearing on whether your data is in the cloud. That is a separate decision, about where the application server runs. The edge approach works both with a server on your own infrastructure and with a cloud instance.
What the term “cloud ANPR” covers
“Cloud ANPR” is used for two different things. Sometimes it means analysis on a server in your own server room, in other words local. Sometimes it means sending images to an external recognition service, which has entirely different consequences for your data and for the link. When you compare offers, ask where the camera image physically goes and who has access to it.
Edge and central ANPR in TraxSystem
TraxSystem supports both variants. Central analysis works on RTSP streams from IP cameras, in memory, without continuous recording. The edge variant is the Trax7Edge device at the gate: it reads the plate locally, sends the server a finished read and has its own remote update, so a new version doesn't need a visit to the gate. In both variants a photo of the vehicle is saved with the visit, but neither records video continuously and neither recognises faces.
Number plate recognition is part of the system's base package and is not a separately charged module. Edge devices are hardware and are priced separately.
Questions and answers
Can both variants be mixed on one site?
Yes. A remote gate can work with an edge device and a gate next to the server room with central analysis, and events from both go into the same entry log.
What happens to events from an edge device when the link drops?
The device buffers them locally and sends them to the server once the connection is restored.
Will an edge device open the barrier on its own when the server is unavailable?
That depends on the site's configuration and on how the fallback behaviour was designed. What should happen when the server is unavailable is decided during the pre-implementation analysis, because the system has no default setting for this situation.
See also: What is ANPR: number plate recognition explained · On-premises or cloud: choosing a server for vehicle entry control · Do you need new barriers and cameras to implement the system