traffic analytics

Vehicle detection across parking areas and monitored locations

In short

Ayonix vehicle detection identifies vehicles in parking areas and monitored locations from an existing camera view, and tracks them between frames. It is the primitive beneath vehicle counting, parking occupancy and vehicle zone rules, and reports position and class without reading plates.

Request a pilot Send a sample video

Conceptual illustration of a monitored parking facility. AI-generated conceptual image, not a customer deployment.

Scope

What this analytic does, and what it does not

What it detects

Vehicles present in the camera field of view
Continuous vehicle tracks as they move through the scene
Vehicle presence within a defined zone or bay
Stationary versus moving state, where the view supports it

What it does not guarantee

Identification of a specific vehicle or its registered keeper
Reading of a licence plate — that is a separate ANPR capability with different camera requirements
Detection of heavily occluded vehicles in a densely packed yard
Reliable make, model or colour attribution

The problem

What customers are actually dealing with

These are the situations that lead teams to look at this analytic in the first place.

A yard has cameras but no data on how many vehicles are present or how long they stay.
Barrier and loop counts cover the entrance but not what happens inside the site.
Parking areas are managed by patrol because there is no live view of which areas are full.
Motion-based rules on vehicle areas fire on every passing headlight and shadow.

How it works

Detection workflow

Every event carries the rule that produced it, so an operator can see why they were alerted.

1

Ingest

Camera streams covering vehicle areas are read and monitored for availability.
2

Detect

Vehicles in frame are detected with a confidence value per detection.
3

Filter

Size and confidence thresholds and region masks remove implausible detections and irrelevant areas.
4

Track

Detections are associated into tracks so that one vehicle produces one continuous object.
5

Expose

Tracks and zone membership are made available to counting, occupancy and zone rules.

Configuration options

Confidence threshold per camera, set during commissioning
Minimum and maximum vehicle size in pixels for the covered range
Region masks to exclude a public road or neighbouring site in view
Bay or zone polygons where per-space occupancy is required
Track continuity settings across brief occlusion by other vehicles
Stationary threshold before a vehicle is treated as parked

Alert quality

How irrelevant and duplicate alerts are reduced

No video analytic eliminates false alerts. These are the mechanisms that reduce them, and the residual rate is measured on your own cameras.

Tracking means a vehicle driving through the scene produces one object rather than a detection per frame.
Region masks remove public roads and adjacent property, which are the commonest source of irrelevant detections.
A stationary threshold prevents a vehicle slowing at a junction from being recorded as parked.
Size filtering separates vehicles from large objects and from distant traffic outside the area of interest.
Per-camera behaviour varies with angle and range and is characterised during the readiness review.

Prerequisites

Camera requirements and environmental limits

Camera suitability decides more of the outcome than any software setting. These are assessed per camera before commitment.

Camera requirements

A view where the vehicle occupies a meaningful part of the frame at the furthest bay or lane of interest
An elevated angle that limits one vehicle occluding another, which matters most in packed yards
Adequate illumination for the hours of operation, including IR where the area is unlit
Stable mounting so bay polygons stay aligned with the physical markings
Exposure settings that cope with headlights at night without blooming across the frame

Environmental limitations

In densely packed areas, vehicles occlude each other and detection of the rear rows degrades.
Strong headlight glare at night can wash out detail across part of the frame.
Heavy rain, snow cover and standing water change vehicle appearance and reduce reliability.
Very oblique views compress distant bays to the point where per-bay resolution is not achievable.

Delivery

Alerts, evidence and where they land

Event and evidence fields

Track identifier and object class
Detection confidence
Zone or bay membership
Stationary state and duration
Camera and site
First and last seen timestamps

VMS integration

Vehicle detection is the input to zone and counting rules rather than a standalone alarm. Vehicle zone events, such as presence in a restricted yard area outside hours, are delivered to the VMS in the same way as person-based zone events.

See compatibility states →

Deployment options

Edge processing at the yard or car park, keeping video local
On-premise processing where site cameras already aggregate to a local server
Multi-site deployment for consistent yard measurement across depots

Compare architectures →

Measurement

Dashboard metrics and pilot acceptance criteria

Acceptance thresholds are agreed with you before the pilot starts. This site publishes no benchmark figures, because they do not transfer between sites.

Dashboard metrics

Vehicle detections and tracks per camera per hour
Occupied bays or zones over time
Mean stationary duration per zone
Confidence distribution per camera
Cameras below the readiness threshold for vehicle rules
Stream availability

Measurable pilot criteria

Detection recall across near, mid and far bays under day and night conditions
Behaviour with a full yard, where occlusion is at its worst
False detections per camera per day on an empty area
Stationary-state accuracy against observed arrival and departure times
Per-camera pass or fail against the agreed readiness threshold

Governance

Privacy and governance

Detection identifies a vehicle as an object. It does not read plates and does not identify a keeper or driver.
Where a camera view includes a public road, region masking should be used to limit processing to the site.
Vehicle tracks are scoped to a camera session and are not a persistent vehicle identifier.

Frequently asked questions

Does vehicle detection read number plates?

No. It detects and tracks vehicles as objects. Plate reading is ANPR, which has substantially stricter camera, angle, shutter and lighting requirements and its own privacy considerations. Many sites find vehicle presence answers the operational question without plates.

Will it work in a full car park?

The front rows will detect well; rear rows in a very oblique view may not, because vehicles occlude each other. Camera height and angle are the determining factors, and the readiness review reports per-camera coverage rather than assuming the whole area works.

Can it tell a van from a car?

Broad class separation is possible in views with adequate resolution, but reliable make, model and detailed classification are not claimed here. Where class granularity matters commercially, it is characterised specifically during the pilot.

Keep reading

A controlled pilot establishes what this analytic actually does in your environment, against acceptance criteria we agree before it starts. Send a sample video first if you would rather see an assessment before committing to a pilot.