AI that detects, alerts and speaks up

AI analytics inside the NVR that finds people, vehicles, fire and plates and instantly flags intrusion, missing hard hats and wrong-way driving.

5
Detection targets: 7
Event rule types: 360°
Fisheye dewarped for analysis

Talk to sales Download catalog

Ubuntu 24.04 LTSWindows 10 / 11RK3588 (ARM)

Rule setup — plate recognition and thermal options

Highlights

What it does

Down to hard hats and wrong-way

Draw a zone or line and monitoring for intrusion, falls, wrong-way driving or missing hard hats starts right away. Each rule can also speak an announcement through the NVR speaker.

People, vehicles and fire, live

Detects and tracks people, vehicles, faces, fire and plates on the device, and saves the moment a rule fires as a best shot.

VCA viewer — live detections and best-shot panel

Find it, play it

Find best shots by type, channel or plate number and play the recording from that moment.

Best-shot search with playback at the event time

Straight to the control room

Events go out over HTTP and MQTT to the NQVMS e-map, NQMobile alerts and external control systems.

Features

Features at a glance

Five detection targetsPerson, vehicle, face, fire and plate; enable only what each channel needs.
Seven rule typesDetection, intrusion, counting, plate, fall, wrong-way and hard hat, applied without a restart.
Hard-hat detectionChecks workers in the zone and alerts once per person without a hard hat.
Voice announcementsReads a message through the NVR speaker; phrase and repeats are set per rule.
Fisheye dewarped analysisDewarps ceiling fisheye video before analysis so people and hard hats are not missed.
Korean plate recognitionA dedicated Korean plate model with format checks cuts misreads and duplicates.
Adjustable confidenceTune the threshold per person, vehicle, face and fire to balance false alarms and misses.
Automatic CPU/NPU selectionUses whichever is installed: Hailo-8, RK3588 NPU or OpenVINO.

Screens

SpecificationsShow full specifications

Product description

NQVCA is the AI analytics built into NQNVR: detection, tracking and rule evaluation run on the device with no cloud. It uses whichever accelerator is installed, Intel CPU (OpenVINO), Hailo-8 or RK3588 NPU, and results are kept as best shots and events and forwarded over HTTP and MQTT to control systems. Fisheye cameras are dewarped before analysis, so even ceiling cameras catch people and hard hats.

NQVCA is an optional NQNVR component; available features follow the VCA / LPR license.

Analytics
Detection targets	Person · vehicle · face · fire · license plate
Further analysis	Korean plate reading · hard-hat check
Rules	Detection · intrusion · counting · plate · fall · wrong-way · hard hat
Tracking	Multi-object tracking with track IDs
Sensitivity and confidence	Per-channel sensitivity and rate, per-target confidence
Voice announcements	Phrase and repeats per rule, NVR PC speaker
Fisheye (360°)	Dewarped before analysis (ceiling or wall mount)
Thermal	Thermal camera temperature alarms

Platform
Inference	OpenVINO (Intel CPU) · Hailo-8 NPU · RKNN (RK3588)
Analyzed channels	Up to 16 of the NVR's cameras analyzed at once
Outputs	NVR events · best shots · HTTP POST · MQTT
Tools	VCA viewer · best-shot search · count statistics · inference benchmark
Applications	NQPlateExtractor · NQVehicleTracker · NQParkingEnforcer
Licensing	VCA / LPR features in the NQNVR license