Home  /  Products  /  AI Edge Computing

Edge AI Video Analytics & AI CCTV Analytics — Turn Existing CCTV into Smart CCTV

Add intelligent video analytics to the cameras you already own. ADHOOPU combines edge AI boxes, AI compute servers and an algorithm management platform into an end-to-edge-to-centre architecture — broad AI algorithm library, multi-algorithm per channel, and no camera replacement. Run it as edge AI video analytics on site, as on-premise AI video analytics on your own server, or both from one platform; where a zone has no coverage at all, AI CCTV cameras can be added only where they are needed.

AI EDGE COMPUTINGAI VIDEO ANALYTICSEXISTING CCTV AI UPGRADE EDGE AI BOXAI COMPUTE SERVERINDUSTRIAL SAFETY AI

AI EDGE COMPUTING BOX · EDGE AI BOX · AI VIDEO ANALYTICS BOX · VIDEO ANALYTICS APPLIANCE · EDGE AI APPLIANCE · AI VIDEO ANALYTICS SERVER · AI VIDEO INTELLIGENCE PLATFORM · EXISTING CCTV AI UPGRADE · EDGE AI VIDEO ANALYTICS · AI CCTV ANALYTICS · AI CCTV CAMERA · ON-PREMISE AI VIDEO ANALYTICS

Three-Layer Architecture: Edge, Centre, Platform

A modern AI video analytics deployment is not one box. ADHOOPU separates the workload across three layers so that projects scale from a single site to a multi-hundred-camera programme without redesign.

Edge Layer — AI Edge Computing Box

QG-615AI16 sits next to the cameras and analyses locally. Up to 16 channels of 1080P, real-time alarms on site, and drastically less bandwidth back to the centre because only events travel the network.

  • Edge AI box for distributed sites
  • 16-channel local AI analytics
  • Low power, fanless operation
  • Works with existing IP cameras

Centre Layer — AI Compute Server

QG-6181U96 / 6182U128 / 6184U256 centralise analytics for large programmes: 96, 128 or 256 channels of 1080P analysis in 1U, 2U or 4U, with optional large-model re-checking to suppress false alarms.

  • AI video analytics server for central aggregation
  • 96 / 128 / 256 channel models
  • Multi-channel AI video analytics
  • Large-model re-check option

Platform Layer — Management & Algorithm Platform

The AI video intelligence platform manages devices, channels, algorithms and alarms, and controls how the algorithm set evolves over the life of the project.

  • AI algorithm management platform
  • AI model management and remote AI model upgrade
  • Edge AI node management
  • Alarm workflow and reporting

Edge AI Video Analytics — Algorithm Coverage (a broad AI algorithm library)

This is the complete verified algorithm set. The same library runs on the edge AI box and the AI compute server, with multi-algorithm parallelism per channel (10+ algorithms concurrently on a single video stream).

Personnel Compliance (8)

  • Helmet detection — no safety helmet
  • Mask detection — no mask
  • Workwear detection — no workwear
  • Safety belt detection — no safety belt
  • Reflective vest detection — no vest
  • Riding without a helmet
  • Motorcycle entering a fuel station
  • No workwear in production zones

Behaviour Warning (12)

  • Fall detection
  • Smoking detection
  • Phone use detection — calling
  • Mobile phone detection — looking at phone
  • Running detection
  • Sleeping on duty / sleep on duty detection
  • Personnel absence detection — leaving post
  • Crowd gathering detection
  • Fight detection / violence detection
  • Personnel loitering / personnel overstay detection
  • SOS gesture detection — hand raise for help
  • Fatigue driving detection

Perimeter Protection (18)

  • Vehicle parking detection — no-parking zone
  • Vehicle leaving
  • Person loitering
  • Wall climbing detection
  • Intrusion detection
  • Line crossing detection — boundary crossing
  • Over-occupancy / under-occupancy
  • Person leaving
  • Non-motor vehicle no-parking / leaving
  • Vehicle over-count / under-count
  • Hazardous vehicle detection — restricted entry
  • Fuel unloading non-compliance
  • Safety sign recognition
  • Vehicle overspeed
  • Vehicle wrong-way driving

Environmental Anomaly (9)

  • Fire detection AI — flame alarm
  • Smoke detection AI
  • Fire equipment detection
  • Debris detection / garbage detection
  • Camera obstruction
  • Camera displacement
  • Leak and spill detection
  • Over-height / over-limit detection
  • Object left behind detection
  • Rat / bird / cat / dog / animal detection

Full-Structured Analytics (5)

  • Face detection / face capture
  • Person detection / human detection
  • Vehicle detection / motor vehicle detection
  • Non-motor vehicle detection
  • License plate recognition (LPR / ANPR)
  • Attribute output for all five object classes
  • Structured video analytics with person-vehicle-plate association

Face & Plate Binding / Recognition (3)

  • Face recognition with comparison
  • License plate recognition with comparison
  • Face recognition alarm on watchlist match
  • Whitelist face recognition
  • Face person association
  • Vehicle license plate association
  • Person non-motor vehicle association

Industrial / Special (5)

  • Working-zone sign detection
  • Street-vendor / illegal display detection
  • Weapon detection
  • Machine running-state detection
  • Forklift detection

Verification Accuracy

  • Face capture ≥99%, misidentification <0.5%
  • Whitelist recognition >99.5%
  • Face database up to 300,000
  • Human body ≥95%, non-motor vehicle ≥95%
  • License plate recognition ≥95%
  • Motor vehicle ≥90%, other alarms ≥90%
  • False capture rate <1%

Why Upgrade Existing CCTV Instead of Replacing It

Most sites already have cameras that work. The problem is not image quality — it is that nobody is watching them at 3am. AI analytics converts the existing video stream into events.

Protect Existing Investment

You keep your current IP cameras, NVR and VMS. Add an AI edge box or AI compute server, and the same cameras start producing events instead of just recordings.

Works with Major Brands

Streams are pulled over ONVIF, RTSP, GB28181 and RTMP, so the upgrade works alongside Hikvision, Dahua and other mainstream camera brands rather than locking you into one vendor.

Phased Rollout, Lower Cost

Because the analytics layer is added separately, sites can upgrade progressively — priority zones first — instead of replacing an entire camera estate in one budget cycle.

Fewer False Alarms

Large-model re-checking and confidence-threshold tuning suppress the false triggers that make traditional motion detection useless in outdoor and industrial environments.

Real-Time Alerts, Not Post-Mortem

The shift is from reviewing footage after an incident to receiving an alarm while it is happening — with the 20 seconds of video before and after the event already captured.

Open Integration

Alarms reach your platform over HTTP, MQTT, WebSocket and REST API, so results can drive existing VMS, access control or building systems instead of sitting in a separate silo.

Product Line

Choose the layer that matches the deployment. All models are available under your own brand through the OEM/ODM programme.

QG-615AI16 — Edge AI Box

16-channel AI edge computing box for local analysis at distributed sites, branch offices, industrial plant and remote locations.

  • 16 channels of 1080P
  • broad AI algorithm library, multi-algorithm per channel
  • Fanless, low-power design
  • Standalone or managed by the central server

View the 615 AI Box →

QG-618xU — AI Compute Server

AI video analytics server for centralised programmes: 96, 128 or 256 channels of 1080P in 1U, 2U or 4U.

  • 96 / 128 / 256 channel models
  • broad AI algorithm library
  • Optional large-model re-checking
  • 2 to 22 edge compute units per chassis

View the AI Server Line →

Existing CCTV AI Upgrade

The most requested deployment: add analytics to cameras that are already installed, without replacing them.

  • No camera replacement, no re-cabling
  • Works with Hikvision, Dahua, others
  • Phased rollout by priority zone
  • Alarms into your existing VMS

See the Upgrade Guide →

AI Edge Computing FAQ

What is AI video analytics?
AI video analytics (also called AI CCTV analytics) uses deep-learning models to interpret live camera streams instead of only recording them. Each channel is analysed for defined objects and behaviours, and an event is raised when something matches. ADHOOPU ships broad AI algorithm library either as an edge AI video analytics appliance on site or as on-premise AI video analytics on a central server — with no camera replacement.
Edge AI video analytics or on-premise AI video analytics — which should I choose?
Choose edge AI video analytics when bandwidth, latency or site autonomy matter: the box sits in the local network, analyses up to 16 channels and returns alarms. Choose on-premise AI video analytics on an AI video analytics server when the camera count runs into the hundreds and you want one central algorithm library, cross-camera search and large-model re-checking. Multi-site projects usually run both and manage them from one platform.
Do I have to replace my existing cameras to get AI analytics?
No. This is the entire purpose of the existing CCTV AI upgrade architecture. ADHOOPU analytics boxes and servers pull the video stream from your current cameras over ONVIF, RTSP, GB28181 or RTMP and run the algorithms centrally or at the edge. Your cameras keep recording as before — they simply start generating events as well.
How many algorithms can run on one video channel at the same time?
The ADHOOPU edge AI box supports more than 10 algorithms concurrently per channel (multi-algorithm video analytics). Different channels can run different algorithm sets depending on the scene — a perimeter camera and a PPE-compliance camera use the same box with different configurations.
What is the difference between the edge box and the AI server?
The 615 AI edge computing box handles up to 16 channels locally at the site — ideal for distributed locations where you want alarms on site and minimal bandwidth back to the centre. The AI compute server (96 / 128 / 256 channels) centralises analysis for large programmes and adds optional large-model re-checking. Many projects use both: edge boxes at remote sites, servers at the centre.
Can the AI box integrate with our existing VMS or platform?
Yes. Alarms and structured metadata can be delivered over HTTP, MQTT, WebSocket and REST API, and results can be pushed as re-encoded AI video streams to third-party platforms. AI video analytics API and AI analytics SDK integration are available to partners.
How accurate are the AI algorithms?
Verified engineering figures: face capture ≥99% with whitelist recognition >99.5% and misidentification <0.5%; human body, non-motor vehicle and license plate ≥95%; motor vehicle ≥90%; other alarms ≥90% — all with false capture rate <1%. Industrial safety algorithms are tuned for hazardous-chemical and production environments.
Is the AI analytics hardware available for OEM / private label?
Yes. ADHOOPU manufactures AI edge box OEM and AI video analytics ODM hardware under partner brands, including private label AI box products, custom algorithm sets, custom GUI and API/SDK integration.

Plan Your AI Video Analytics Upgrade

Tell us how many cameras you have, whether they are already IP-based, and which behaviours you need to detect. We reply within 24 hours with a recommended architecture and channel count.

Manufacturer Reference

ADHOOPU

Professional AV manufacturer — Integrated AV Platforms, Distributed AV over IP, and AI Edge Computing & Video Analytics. OEM/ODM for system integrators, distributors and AV brands worldwide.

19 years of AV engineering · Shenzhen, China · 3-year warranty

Products

Integrated AV Platform QG-1169 AV over IP Encoders ST52K Encoder SR52K Decoder AI Video Analytics 615 AI Analytics Box

Solutions

Warehouse Safety Industrial Safety Oil & Gas Safety Hospital Safety Campus Security Parking Analytics Traffic Analytics Smart City Analytics Logistics Safety Retail Analytics Airport Analytics Smart Courtroom Surgical Teaching Command Center Education Meeting Room Smart Security Construction Site Factory & Industrial

Company

OEM / ODM About ADHOOPU FAQ Resources Contact
ADHOOPU · Room 1206, Block A, Dachanwan, Jingang Commercial Building, Xixiang Street, Bao'an District, Shenzhen, China
WhatsApp +86 138 2352 5531 · Email jimmy@adhoopu.com · Web https://www.adhoopu.com
© 2026 ADHOOPU. Integrated AV Platform · AV over IP Encoder / Decoder · AI Video Analytics · Professional AV OEM Manufacturer.

AI Video Analytics Appliance: ONVIF and RTSP Camera Integration with On-Premise Edge Inference

The 615 is an AI video analytics appliance, edge AI video analytics box and AI camera analytics box that upgrades existing CCTV into smart surveillance. It provides ONVIF AI video analytics, RTSP video analytics, ONVIF IP camera integration, RTSP camera integration and broad third-party IP camera compatibility — no camera replacement required.

All local AI processing runs as edge inference on the box: cloud-free AI video analytics with no per-camera cloud fees. As a 16-channel AI video analytics box and AI NVR it delivers multi-channel AI video analytics, real-time AI alerts, AI event recording and event-triggered recording with pre- and post-event buffering.

Does the AI box support ONVIF and RTSP cameras?

Yes. The box ingests any ONVIF-conformant IP camera or any RTSP H.264/H.265 stream, so it works with existing CCTV cameras from mainstream manufacturers — a true third-party IP camera compatibility upgrade path.

Is AI video analytics processed locally or in the cloud?

Locally. All inference runs on the edge AI box itself — on-premise AI video analytics with cloud-free operation, keeping video inside your network.

Can the AI system send real-time alerts?

Yes. The box raises real-time AI alerts with event snapshots via HTTP, MQTT and webhook interfaces, and supports event-triggered recording with pre-event and post-event video buffers.