VaultrisPrivileged Access & Traffic Intelligence
Vaultris

Products

Bastion HostGateway, vault, session recording Intelligent TrafficCorridor monitoring, LPR & BOLO

Industries

Military & Defense Police & Law Enforcement Fire & Emergency Services All industries

Resources

Bastion host guide Compliance & evidence Field notes

Company

About Vaultris Partner programme Request a Demo

Intelligent Traffic

Know the road is blocked before the calls start.

Vaultris Intelligent Traffic watches every camera on a toll network, decides when a corridor has actually stopped moving, and tells the right people on the schedule they set. A licence plate recognition module runs alongside it — reading plates in real time, matching them against BOLO records, and identifying vehicles by colour, body type and model.

Request a demonstration See the LPR module

What it does

Four jobs, on one screen.

A traffic operations centre does not need four products. It needs the queue, the cause, the vehicle, and the record — in one place, with one audit trail.

01

Watch the corridor

Every camera is scored continuously for flow, queue length and speed. A segment is called congested from sustained evidence, not a single frame.

02

Tell the right people

Members choose which corridors they care about, which hours they want to hear about them, and which hours they never do. Critical events override both.

03

Read the plate

The LPR module reads plates lane by lane, matches every read against active BOLO records, and notifies the designated party on a hit.

04

Describe the vehicle

Colour, body type and make/model are recognised on the same pass — so a vehicle can be found from a witness description, with no plate at all.

1 sDetection to alert, typical
7Vehicle classes recognised
24/7Unattended operation
0Plates leaving your network

Network Monitoring

One map that shows the whole road, not fourteen tiles.

Camera sites are placed on the corridor they actually belong to, and each segment between them is coloured by the state of the traffic on it. An operator can see at a glance which stretch of the network has stopped, and which is simply busy.

  • Corridor-level rollup — the whole route carries a state, so a single stalled plaza does not read as a network-wide failure.
  • Five states, not a number to interpret: free, light, moderate, heavy, jam — each with its own speed and queue thresholds.
  • Live event counts per site, so a camera that has quietly stopped contributing is obvious.
  • Plate-reader sites are marked — operators know which points can answer "which vehicle", not just "how many".
  • Camera wall on demand with per-lane queue estimates under every feed.
Traffic operations centre with the Vaultris camera wall on the main screen — sixteen toll-road feeds, each carrying its detected congestion state and vehicle queue
The camera wall on the operations floor. Each feed carries the state the detector assigned it and the queue it measured — the operator is not asked to eyeball sixteen videos.

Alerts & Schedules

An alert at 3 a.m. had better be worth waking someone for.

Notification fatigue is how monitoring systems get muted, and a muted system is worse than none. Intelligent Traffic makes the schedule a first-class setting: who hears about what, during which hours, and which hours are silent no matter what.

  • Alerting hours per day of week — a grid, not a single on/off switch. Weekday peaks and weekend patterns are rarely the same.
  • Quiet hours hold non-critical alerts and deliver one digest when the window ends. Nothing is discarded; it is deferred.
  • Block-out windows suppress notification entirely — planned maintenance, resurfacing, a closed plaza.
  • Critical override lets BOLO matches and wrong-way detections break through both windows. Safety events are never held.
  • Re-alert cool-down stops one long jam from generating forty messages.
  • Per-rule routing — SMS, voice call, email, ops console or webhook, chosen per rule and per recipient group.
Alert settings screen with a seven-day by twenty-four-hour grid marking active alerting hours, quiet hours and block-out windows, beside a list of notification rules
Alerting hours, quiet hours and block-outs are set on one grid. The rules that must ignore all of it are marked explicitly.

Members, not just operators. Alerts are not limited to the control room. Corridor notices can be routed to maintenance crews, enforcement liaisons, fleet operators, or a public subscriber list — each with its own corridors, hours and delivery channels.

LPR Module

Plates read in real time, at the lane.

Recognition runs at the camera, on the vehicle as it passes. Reads are timestamped, tied to a lane and direction, scored for confidence, and matched against the watchlist before the vehicle has left the plaza.

LPR command centre showing a live lane view with the identified vehicle boxed and its plate, colour, body type and model read out beneath, beside a scrolling list of recent plate reads
The LPR command centre. The read, the lane, the direction, the confidence and the vehicle description arrive together — a watchlist hit is flagged in the same list.

Camera imagery on this page is licence-cleared photography carrying the product’s own detection overlay — see image credits. Every plate, vehicle record, alert and figure shown is invented sample data.

Read at the edge

Inference runs on the camera or a local appliance. Only the resulting read leaves the lane, so the system keeps working when the link to the centre does not.

Confidence, not certainty

Every read carries a score. Low-confidence reads are held below a configurable floor rather than presented as fact, and the floor is yours to set.

Lane and direction

A read without a lane and a heading is hard to act on. Both are recorded, which is what makes wrong-way and reversing detection possible at all.

BOLO Matching

Put a plate on the list. The network watches for it.

A BOLO — be-on-the-lookout — record ties a plate to a reason, a priority, and the party who must be told when it appears. Every read on every reader is matched against every active record, and the designated recipient is notified on a hit.

  • Exact or partial plates — a partial record still matches when only some characters are known.
  • Categorised and prioritised: stolen, wanted, Amber Alert, parole, DUI warrant, expired registration, lapsed insurance, other.
  • Named notification party per record, so the alert reaches the officer or agency that raised it — not a shared inbox.
  • Match history against each record: where it was last seen, when, and at which reader.
  • Priority drives the channel — a high-priority record can escalate to voice call and ignore quiet hours.
  • Every issue, edit and deletion is logged against the person who made it.
BOLO watchlist showing six active records, each with plate, priority, category, vehicle description, match mode, issuing officer and last sighting
The watchlist. Each record states its category, who issued it, how it matches, and where the plate was last seen.
A BOLO match alert showing the matched plate, the reason, the reading camera and lane, the confidence score, and the officer notified
What a match looks like on the operator's screen: the plate, the reason, the reader, the confidence, and who has already been told.

Plate Tracking

Two readers make a journey. One makes a sighting.

When the same plate is read at more than one point, the route between those points is known — and it is drawn along the road network, not as a straight line between pins. Direction, elapsed time and the distance actually travelled follow from the corridor geometry.

  • Journeys along the real alignment, so an operator sees the road the vehicle used.
  • Sighting list per plate with reader, lane, time and read confidence at each point.
  • Corridor view of every vehicle seen in the window, filterable to journeys only.
  • Exportable as a record for a case file, with the confidence of each read preserved.
Realtime tracking screen following one plate along the Turnpike corridor, with the route drawn on the road network and four timestamped sightings listed
One plate, four readers, a route that follows the carriageway. The sighting list carries the confidence of each read.

Vehicle Identification

"Silver SUV, northbound, sometime after eleven."

That is what a witness gives you. It is also a valid query. Colour, body type and make/model are recognised on every read, so a vehicle can be found without a registration number.

Vehicle identification screen showing recent reads with the identified vehicle boxed in each frame, its colour, body type and model listed with separate confidence scores, above colour and body-type distribution charts
Each attribute carries its own confidence. Make and model is the hardest of the three and is reported as such, rather than folded into one number.

Colour

Nine colour families, scored independently of the plate read. Colour survives a dirty or obscured plate, which is exactly when it matters.

Body type

Sedan, SUV, pickup, van, truck, bus, motorcycle. The same classes drive the traffic counts, so a survey and a search agree with each other.

Make and model

Reported with a lower and honestly-stated confidence than colour or body type. Useful for narrowing a search; not offered as an identification on its own.

Counts & Reports

The same detections that raise alerts also do the counting.

Congestion monitoring and traffic surveying are the same measurement viewed over different windows. Every detection is retained as a count, so classified volumes, turn movements and peak profiles come out of the system already collected.

  • Classified counts by vehicle class over any window, in buckets from one minute upwards.
  • Turn-movement surveys per approach — straight, left and right counted separately, which is what signal timing and queue models need.
  • Peak profiles and busiest-bucket figures, derived rather than estimated.
  • Confidence distribution published alongside the counts, so the quality of the underlying data is visible.
  • CSV and PDF export for modelling and for the record.
Detection analytics showing 48,170 detections over an hour in five-minute buckets, stacked by vehicle class, with class distribution and a read-confidence histogram
Classified detections over an hour. The confidence histogram is shown next to the counts, not buried — the discard floor is visible.
Traffic survey report for one approach showing turn movements over time, movement breakdown percentages and the vehicle-type mix
A turn-movement survey for a single approach. Turning volume is usually the number a signal plan hinges on, and it is counted, not apportioned.

Detection Accuracy

A camera that counts a road sign all day is worse than one that is switched off.

Every detection zone is audited continuously against its own behaviour. A zone that reports constant traffic with no daily profile is almost certainly tagging something stationary. A zone reporting almost nothing is probably pointed at open ground. Both are flagged automatically instead of quietly corrupting the counts.

  • Static-object detection — busy nearly all day, one class dominating, no track diversity.
  • Misplaced-zone detection — too few active hours and too few detections to be a real traffic lane.
  • Hourly activity profile shown per zone, so the diagnosis is visible rather than asserted.
  • Verification against manual counts supported and reported as a variance, not hidden.
Lane queue health table listing detection zones with 24-hour detections, active hours, track rates, dominant class, hourly activity sparklines and an automatic health flag
Zone diagnostics. The sparkline is the evidence for the flag beside it — a flat all-day profile is what "static object" actually looks like.

Data Governance

Plate data is the most sensitive thing on the network.

Licence plate reads describe the movements of identifiable people. Intelligent Traffic is built to be governed the same way privileged access is governed on the rest of the Vaultris platform: least privilege, a retention limit, and a record of every search.

ControlWhat it means in the product
Retention limitPlate reads expire on a configured schedule. Reads attached to an active BOLO match or an open case are held separately and explicitly, rather than by extending retention for everything. Statutory limits vary from minutes to three years depending on the state — see the retention comparison and what each period costs.
Purpose-bound accessRoles separate live monitoring, historical search, BOLO administration and export. An operator watching queues does not need historical plate search, and does not get it.
Searches are auditedEvery plate query, journey reconstruction and export is written to an append-only log against the named user, with the reason recorded. The audit covers who looked, not just who changed something.
Enforced multi-factor authenticationAccess to plate data goes through the same MFA-enforced gateway as the rest of the platform — see the access lifecycle.
Stays on your networkInference runs at the edge and data is held on infrastructure you control, including fully air-gapped deployments. No plate read is sent to a Vaultris-operated service.
Exportable recordReads, matches and the audit log export in full for disclosure, oversight review or a case file.

These are product capabilities, not a compliance opinion. Retention periods, permitted uses and sharing rules for licence plate data are set by your own jurisdiction, agency policy and — for criminal intelligence systems — by regulations such as 28 CFR Part 23. Intelligent Traffic is designed to let you configure and evidence those rules. It does not decide them for you, and nothing here should be read as legal advice.

Deployment

It runs where the road is.

Edge

At the camera

Detection and plate recognition run on the camera or a roadside appliance. A plaza keeps counting and keeps matching BOLOs through a link outage, and forwards what it recorded when the link returns.

Site

In your operations centre

The map, camera wall, watchlist and reporting run on infrastructure you control — on-premises, in your private cloud, or in a sovereign region.

Isolated

Fully air-gapped

No component requires outbound internet access. The same deployment model the Vaultris gateway uses for isolated networks applies here.

Integration

Governed by the platform you already run.

Intelligent Traffic is not a separate island with its own logins. It authenticates through the Vaultris gateway, inherits the same directory, the same enforced MFA, and the same append-only audit ledger — so a plate search and an administrative session appear in one record.

  • Single identity from Active Directory, LDAP, SAML or OIDC.
  • One audit ledger covering access, configuration change and plate search alike.
  • Webhook and CSV outbound into CAD, RMS, ATMS or a variable-message-sign controller.
  • Camera-agnostic — ONVIF and RTSP sources, with no requirement to replace working cameras.

See the underlying platform

webhook — corridor alert
POST /atms/incident  HTTP/1.1
Content-Type: application/json

{
  "event":      "congestion.jam",
  "corridor":   "TPK",
  "segment":    "TPK_GG04",
  "state":      "jam",
  "sustained_s": 248,
  "queue_veh":   18,
  "avg_mph":     11,
  "lanes":      [2, 3],
  "detected_at": "2026-08-19T17:12:44Z",
  "notified":    ["ops-desk", "maint-oncall"]
}

Questions

What agencies ask first.

What is Vaultris Intelligent Traffic used for?

It monitors toll roads and other managed corridors for congestion, notifies chosen members when a segment stops moving, and runs licence plate recognition with BOLO matching alongside the same camera network. It also produces classified traffic counts and turn-movement surveys from the same detections.

How does it decide that a road is congested?

Each camera is scored continuously for vehicle flow, queue length and average speed, and a segment is assigned one of five states — free, light, moderate, heavy or jam. The state must hold for a configured period before it raises an alert, so a single stopped frame or a passing gap in traffic does not trigger a notification.

Can members choose when they are alerted?

Yes. Alerting hours are set per day of week on a 24-hour grid. Quiet hours hold non-critical alerts and deliver them as one digest, and block-out windows suppress notification entirely for planned work. A critical override lets BOLO matches and wrong-way detections break through both.

What does the LPR module do when it matches a BOLO plate?

The match is raised on the operator console within about a second of the read and the party named on that BOLO record is notified through the channels set for its priority — SMS, voice call, email or webhook. The alert carries the plate, the reason, the reader, the lane and direction, the read confidence and the frame. High-priority records ignore quiet hours and block-out windows.

Can it find a vehicle when nobody has the plate?

Often, yes. Colour, body type and make/model are recognised on every read, so a description such as "silver SUV, northbound, between 11:00 and 12:00" is a valid search. Each attribute carries its own confidence; make and model is the least reliable and should be used to narrow candidates rather than to identify a vehicle on its own.

How is a vehicle tracked across the network?

By matching the same plate at more than one reader. Two sightings define a journey, which is drawn along the corridor the vehicle actually used rather than as a straight line, with the elapsed time and distance following from the road geometry. A single reader produces a sighting, not a journey — which is why reader placement matters more than reader count.

Does any plate data leave our network?

No. Inference runs at the edge and reads are held on infrastructure you control. No plate read is sent to a Vaultris-operated service, and no component requires outbound internet access — the system deploys on-premises, in a private cloud, in a sovereign region, or fully air-gapped.

How long are plate reads kept, and who can search them?

Retention is a setting you configure, and reads expire on that schedule; reads tied to an active match or open case are retained explicitly rather than by extending retention for everything. Access is role-separated between live monitoring, historical search, BOLO administration and export, and every search is written to an append-only audit log against the named user with the reason recorded.

Do we have to replace our existing cameras?

Usually not. The system takes ONVIF and RTSP sources, and detection zones are drawn per camera in the interface. Where a camera's placement genuinely cannot support plate recognition — angle, resolution or lighting — the zone health tooling will say so rather than producing unreliable reads.

Who is it built for?

Toll authorities and state transportation departments, law enforcement agencies working those corridors, and emergency management organisations that need a corridor's real state during an evacuation or incident. It is part of the same platform Vaultris builds for military commands, law enforcement and the public safety mission — not a commercial product adapted afterwards.

Next step

See it running on your corridor.

A demonstration uses your camera layout and your corridors, so what you are looking at is your network — not a showcase built to look good.

Request a demonstration Explore the platform