GCXONE · Features
GCX SmartVision
AI Video Analytics for GCX One
Sees the Object. Judges the Behaviour.
GCX SmartVision is the AI video-analytics layer of GCX One. It identifies what is in a scene — people, vehicles, objects — and then applies rule-based decision logic to judge whether what they are doing should raise an alarm. Analysis runs in real time on every incoming event, so an operator is presented with classified, bounded, labelled activity rather than raw motion. Detection is deliberately object-based: no facial recognition or biometric identification is performed at any stage.
Key capabilities
- Two-stage analysis — identification of every object in the scene, then a behavioural decision on whether it matters
- Object classification for people, vehicles and other object types, with confidence levels and bounding boxes
- Configurable decision logic: priority list, white list, black list and reject-unknown, per deployment
- Alarm metadata written onto the event, so an operator verifies a labelled detection rather than a bare frame
- Scheduled analytics, with every active schedule visible in one table across the whole hierarchy
Primary use cases
- Perimeter and outdoor sites where environmental motion, not intrusion, is what generates the volume
- Sites that need vehicle activity and person activity treated as different events with different responses
- Operations that need analytics running on a schedule rather than continuously, by site or by device
- Verification workflows where the operator has to see what the system thought it saw
At a glance
Up to 99%
False alarms reduced
2
Analysis stages
4
Decision list types
3
Detection classes
Decision logic
| Control | Effect |
|---|---|
| Priority List | Objects that always raise a real alarm |
| White List | Objects that may raise an alarm, judged on behaviour |
| Black List | Objects that are always ignored |
| Reject Unknown | Suppresses environmental triggers with no identified object |
Vertical
Smart Security
Intrusion, loitering and perimeter activity on outdoor and unmanned sites.
- Best for: sites where the question is whether a person is there at all
Vertical
Smart Safety
Activity that puts people at risk rather than property — presence in restricted zones, unsafe movement around plant.
- Best for: industrial and construction environments
Vertical
Smart Compliance
Evidence that a rule was followed, captured automatically and timestamped.
- Best for: audited environments where the record is the deliverable
Vertical
Smart Traffic
Vehicle classification and movement, parked versus moving, across car parks and access roads.
- Best for: perimeters and vehicle access control
Vertical
Smart Retail
People movement and occupancy patterns in customer-facing space.
- Best for: operational insight rather than alarms
Vertical
Smart Operations
Activity used as an operational signal — arrivals, dwell, throughput — rather than as a security event.
- Best for: sites where video is a data source, not only a camera
Architecture
Measurable results
Technical specifications
- Stage 1 — Identification
- Detects and identifies every object in the scene: people, vehicles, animals, other objects
- Stage 2 — Decision
- Applies the configured logic to decide whether the event is a real alarm or is suppressed
- Decision Controls
- Priority List (always alarms), White List (alarms based on behaviour), Black List (always ignored), Reject Unknown (filters environmental triggers such as wind and reflections)
- Behavioural Analysis
- Evaluates movement patterns and activity over time rather than a single frame
- Human Detection
- Identifies human figures, draws bounding boxes, reports confidence, tracks movement across the field of view
- Vehicle Detection
- Cars, trucks, motorcycles, buses, bicycles and trains, distinguishing parked from moving
- Object Detection
- Identifies objects that may trigger alarms, supports custom object types, analyses behaviour patterns
- Filtered Conditions
- Lighting changes and shadows, camera vibration or movement, reflections and glare, environmental motion such as trees or rain
- Alarm Quad View
- Pre-Alarm, Alarm, Post-Alarm and Preview frames presented together for every event
- Event Metadata
- Bounding boxes and labels for people and vehicles written onto the alarm
- On-Camera Analytics
- Line crossing, intrusion, human detection and vehicle detection running on the camera or NVR, which is where volume is best controlled
- Measured Effect
- Sites moving from basic motion detection to on-device intelligent detection commonly fall from 200+ alarms a day to under 20
- Analytics Scheduler
- One table of every configured schedule across the hierarchy, showing name, access level, analytic type, schedule time, site and device
- Schedule Levels
- Service Provider, Site or Device — a schedule created at a parent level applies downward
- Duplicate Handling
- Duplicates are listed rather than merged, so a redundant job can be found and removed
- Detection Scope
- Object, person and vehicle detection only. No facial recognition and no biometric identification is performed at any stage
- Camera Dependency
- Analytics quality is bounded by the feed. A camera below the resolution or lighting a model needs produces unreliable results whatever the configuration says
- Not A Replacement
- SmartVision classifies and suppresses. It does not dispatch, does not decide response, and does not remove the operator from the loop
Analysis
Detection types
Operator context
Source detection on the device
Scheduling
Governance & Limitations
Getting started
Subscribed per camera from the Marketplace under Analytics, then scoped and scheduled at whichever level suits the deployment.
- 01Check the feedConfirm resolution, lighting and camera type meet the model requirements before subscribing
- 02Subscribe the camerasMarketplace, then Analytics — activate and pick the cameras
- 03Set the decision logicConfigure the priority, white and black lists and reject-unknown for that environment
- 04Review the scheduleOpen Analytics Scheduler and confirm one schedule per entity, with no duplicates
Works with
