GCXONE · Fonctionnalités

NOVA99x

AI False-Alarm Filtering Engine

Filters the Noise. Trusts the Signal.

NOVA99x is an intelligent false-alarm filtering module for NXGEN’s GCXONE platform. It processes motion- and analytics-triggered events in real time and uses computer vision to classify each event as a true alarm or a false alarm before it reaches an operator, filtering up to 99% of non-threatening activity before human review. NOVA99x is licensed per camera stream.

Fiche technique · v1.0 · vérifiée 10 September 2026

Fonctions clés

Removes repetitive, low-value alarms to reduce alert fatigue
Scales monitoring operations without adding headcount
Object detection tuned for outdoor perimeter and site scenes
Tracking and Re-Identification (ReID) filters non-moving objects like parked vehicles

Cas d’usage principaux

Consistent noise control across all of a provider’s customers
Customer-level alarm policies matched to each environment’s risk tolerance
Stabilizes alarm volume during large-scale customer or site onboarding
Protecting SLA-backed alarm response and resolution times

Performance at a glance

Up to 99%

Non-threats filtered

+2 hrs

Hours saved per shift (HESTA)

2.3x

Cameras per operator (HESTA)

Configuration levels

Camera type

Optical (RGB)

Minimum resolution 640×480 pixels
Lighting of at least 10 lux, or ≥30% average image intensity
A person should occupy ≥300 px² (≥0.3% of image area) for reliable detection

Camera type

Thermal

Must be configured as Thermal in GCXONE for the correct model
A detected object should occupy ≥600 px²
White-hot or black-cold configuration required; other palettes cause false negatives

Getting started

Activated through the GCXONE Marketplace and takes effect immediately for incoming alarms.

01SubscribeActivate NOVA99x from the GCXONE Marketplace (Analytics → Explore → Proceed → Yes)
02Choose scopeApply filtering at Service Provider, Customer, Site, Device, or Sensor level
03Tune detectionSet detection confidence and camera type, optical or thermal, per deployment
04Monitor performanceReview true versus false alarm trends from the NOVA99x dashboard

En bref

Detects: People and vehicles
Filters Out: Animal motion, vegetation, light and shadow changes, headlights, glare and reflections, camera vibration, spider webs, insects, and lens smudging
Not Supported: Fisheye lenses; pan-tilt-zoom (PTZ) or dome cameras that move the scene; dual-sensor thermal/optical cameras, where each sensor must be set up separately; and cameras that zoom onto moving objects
Applies At: Service Provider, Customer, Site, Device, or Sensor level
Fail-Safe Handling: Alarms that cannot be verified in time are still treated as real and sent to the operator
Scope & Limitations: Filters incoming motion-triggered alarms only; does not replace detection analytics, camera hardware, physical security measures, human judgment, live monitoring, or compliance rules. NOVA99x performs object detection and person re-identification only; no facial recognition or biometric identification is performed at any stage. People near the edge of a camera’s frame may not be reliably detected and could be filtered out as non-threatening activity.
Camera Resolution: Analog cameras connected via ADPRO are often capped at CIF (352×288), below NOVA99x’s 640×480 minimum, increasing missed-detection risk in low light. If a lower-resolution feed is unavoidable, upscale it with a standard bilinear scaler and avoid heavy JPEG compression.