Tracking vehicles on a plant site: plate, time on site, lists and the ramp queue
Author: DuoVision team
For: heads of security, logistics and production, directors of plants and warehouse complexes
Summary
- Vehicle tracking on a plant site is built from log rows: plate, entry time, exit time, camera and photo. From those rows you get the time a vehicle spent on site and lists of known, unknown and expected vehicles.
- An event that needs a decision (an unknown plate, a vehicle that stayed too long, an exit with no booking) arrives as a card in the panel. A person decides, not the software.
- Matching against a booking in your accounting system is possible only when that system can hand over data (an export, a database or an API). You find that out on site before launch.
- A dedicated camera on the entry lane reads the plate. Dahua's guide asks for at least 130 pixels of plate width (150 for several countries) and an angle of no more than 30° to the direction of travel.
- For example, AVTONOMER AI cloud recognition is listed at UAH 3,097 to 8,850 per camera per month depending on the plan, while a May 2026 news item quoted UAH 1,549-4,425. Camera and installation are extra.
For: heads of security, logistics and production, directors of plants and warehouse complexes
What vehicle tracking on site is, and how it differs from a barrier
What follows is a scheme we tune to each site: the company defines its own lists, cards and queue. A barrier answers "let it in or not" in one moment. Tracking answers questions that come later: who was on the plant last Tuesday night, how long a truck stood at the ramp gate, whether the vehicle that came in this morning has left. For that the system keeps a log, and every vehicle leaves a trace in it.
The search query "control of truck entry" is mostly about that log. A guard at the checkpoint writes a plate by hand. Software records every vehicle whose plate it reads, at any hour, and keeps the frame. A month later you do not need to recall or recount anything, you open the rows for the period.
What goes into a log row
A minimum set for a plant:
- the plate as the system read it;
- date and time of entry and, as a separate row, of exit;
- the camera that saw it (gate, weighbridge, ramp door, exit);
- a photo of the frame with the plate.
Time on site comes from these fields: from the first row at the gate to the row at the exit. If cameras also cover the weighbridge and the ramp doors, one plate appears in several rows and the differences give the time at each stage. We covered how that time turns into money in the article on losses at the ramp. The plate itself, matching to weighing tickets and recognition prices are in the article on licence plates at warehouse and elevator gates, and we do not repeat them here.
Lists: known, unknown, expected
Without lists the log is an archive. With lists it starts to point at what to look at.
| List | What is in it | What happens when a plate matches or does not |
|---|---|---|
| Known | Company vehicles, regular carriers, employee cars with a pass | The row is written with no alert. A card appears when a known plate shows up at an unusual time or in a zone that is not its own |
| Expected | Plates from today's bookings: who, when, which ramp | A plate from a booking is matched and marked as expected. A booking with no vehicle in the agreed window is also a reason for a card |
| Unknown | Everything not in the two lists above, and plates you entered yourself as unwanted | A card in the panel with a photo and the camera. A person decides: fine, needs checking, needs security |
Market services use the same idea of white and black lists with alerts. For example, AVTONOMER AI lists list management for access control and alerts when specific plates appear.
The card in the panel: what a person sees and decides
Nobody reads a log of thousands of rows, so only the records that need a decision are useful. Such a record arrives as a card: plate, time, camera, photo and the reason the card appeared. Typical reasons:
- the plate is in none of the lists;
- the vehicle has been on site longer than the time set for it;
- an exit with no recorded entry, or an entry with no exit by the end of the shift;
- a plate from a booking did not show up in the agreed window.
The responsible person at the company opens the card, looks at the frame and decides. The system does not block the vehicle and does not accuse the driver. It shows what differs from a normal day. A misread plate (dirty, bent, washed out by headlights) is also rejected by the responsible person on the card. That is how we work with other rules too: the software prepares the card, a person makes the call.
Time limits (how many hours are normal for unloading, how many for waiting) are set by the company. They come from your procedure, not from the software. The first weeks in shadow mode show real figures, and the limits are set from them.
The ramp queue
Trucks arrive at different times and ramps do not free up at once. If a gate camera records the time of every vehicle and a camera at the ramp door shows when it was backed in, a simple queue appears: who arrived first, who is still waiting, for how long. This does not replace a dock booking system that issues slots. It checks it: does the order on site match the order in the bookings.
In practice this gives you:
- a dispatcher who sees who has waited longest without going out to the yard;
- a dispute with a carrier about when exactly the truck arrived settled by a row with time and a frame;
- a view of where the queue forms: at the gate, at the weighbridge or at the ramp.
Accuracy depends on having a camera at each point. Without a camera at the ramp you only know that the vehicle came onto the site.
Matching against a booking in your accounting system
A plate in the log can be compared with a booking: carrier, counterparty, cargo, arrival window. This is a possibility, not always a ready feature. It works when your accounting system (1C, BAS or another) can hand over bookings: as a file export, through a shared database or an API. Whether it can depends on the specific program and its vendor, so you find out before buying, not after.
If the data is available, the system flags mismatches: a vehicle with no booking, a booking with no vehicle, a plate on the booking that differs from the one at the gate. If it is not available, you keep manually maintained lists and the log still works.
Which cameras you need
A plate is read by a camera aimed at the lane, not by an overview camera over the yard. We rely on Dahua's "LPR Camera Selection & Installation Guide v1.1". Figures for other makers may differ, so check them against the documentation of the camera you choose.
| Parameter | What the Dahua guide says | What it means for a plant |
|---|---|---|
| Plate size in frame | At least 130 pixels wide for a one-line plate, 150 for several countries | Check on live video how many pixels the plate takes at the distance you need. Leave a margin |
| Angle | No more than 30° horizontally and vertically to the direction of travel | The camera goes along the lane, not to the side on a pole |
| Speed | For fast traffic shorten exposure to roughly 2.5-3 ms | At a checkpoint where vehicles stop it matters less, on a drive-through without stopping it matters more |
| Night and headlights | Infrared lighting is needed; if the plate washes out, reduce the light power or the exposure | Test at night and under headlights, not only by day under a clear sky |
In practice: one camera per entry lane and one per exit lane, plus cameras wherever you need time per stage (weighbridge, ramp doors). How many you need is clear only on site after a survey.
What cloud plate recognition costs: one service as an example
We do not quote prices for our services because they depend on the site. Public prices for cloud recognition exist. Here is one service, AVTONOMER AI, in two sources as of 03.11.2026.
- The avtonomer.org page: STREAM BASE about UAH 3,097, STREAM PRO about UAH 5,531, STREAM MAX about UAH 8,850 per camera per month. The page converts from dollars at 44.25 UAH, so the hryvnia sums are approximate.
- A dev.ua news item of 01.05.2026 on the launch of AVTONOMER AI gives tariffs from UAH 1,549 to 4,425 per camera per month, depending on archive depth and analytics detail.
The two pages differ by roughly a factor of two. We do not know why: tariffs may have changed after launch. Ask the vendor for current prices before deciding. These sums do not include the camera, lighting, installation or connection to your accounting system.
What vehicle tracking does not give you
- The weight of a load or what is in the body. A plate says a vehicle was there, not what it carried.
- The identity of the driver. A plate points to a vehicle, and we do not offer face recognition.
- A 100% result. A dirty or covered plate, snow or headlight glare give errors or misses, so you need a queue for manual checks.
- Protection against a false plate on a vehicle. The system records the one it sees.
A plate can point to the vehicle's owner, so before launch decide who has access to the log and how long it is kept. General questions are covered in the article video analytics and personal data.
Where to start
- Draw a vehicle's path from the gate to the exit and mark the points where you need to see it.
- Define three lists: who is known, where expected vehicles come from, whom you consider unwanted.
- Ask the vendor of your accounting system whether it can hand over bookings.
- Write down which situations should become cards and who handles them.
- Run the system in shadow mode: only you see the result, and staff get no alerts. After a few weeks set time limits from the real figures.
Plate recognition is among the things we can do. At the meat processing plant where we ran the system, we checked on real events the cap or hairnet, mask, beard cover, gloves, a box directly on the floor and riding a pallet jack. We tune vehicle tracking to each site and start in shadow mode.
Where to apply this
- Warehousing and logistics
- Loading docks and logistics
- Site security and vehicles
- Meat processing
- Case study: video hygiene control at a meat processing plant, 1,192 cards in 8 weeks
- Hard hat detection by camera: construction, warehouse and production. What it sees and where the limits are
- Work clothing control in food production: when work clothes leave the workshop
Sources
- AVTONOMER. Cloud plate recognition, STREAM BASE / PRO / MAX tariffs (checked 03.11.2026)
- dev.ua. A Ukrainian developer launched the cloud service AVTONOMER AI (01.05.2026, tariffs UAH 1,549-4,425 per camera per month)
- Dahua. LPR Camera Selection & Installation Guide v1.1 (ITC237, ITC206, ITC302, ITC602)