Campus Surveillance Network.
AI-Powered Multi-Camera Surveillance.
64–256 camera AI surveillance architecture with perimeter intrusion detection, face recognition, crowd analytics, and 30-day on-premises retention. Designed for corporate campuses, educational institutions, government buildings, and critical infrastructure.
Campus Surveillance Network — System Architecture
A distributed AI-NVR architecture where every 8-camera cluster is served by a dedicated edge node running real-time analytics — face detection, perimeter intrusion, crowd density, and anomaly alerting. A central AI hub aggregates alerts, manages recordings, and provides a unified SOC dashboard. All video processing and storage stays on-premises for full data sovereignty.
How the tiers connect
Tier 1 — Camera Clusters
Every 8 cameras connect to a dedicated AI-NVR edge node via PoE. The edge node runs all AI analytics locally — face detection, intrusion alerts, crowd counting — and forwards only metadata and alert clips to the central hub. Full-resolution recording stays local.
Tier 2 — Central AI Hub
The central hub aggregates alerts from all edge nodes, enables cross-camera person tracking, provides forensic video search, and drives the SOC dashboard. Dual-hub HA ensures zero single point of failure.
Tier 3 — Storage & SOC
RAID-6 storage servers provide 30-day retention with hot-swap drives. The SOC workstation cluster enables real-time monitoring, alert management, and incident response with video wall integration.
Complete BOM — every component specified
All components are RDP-validated, GeM-listed, and available as a single-vendor procurement package. Quantities shown are for a typical deployment — contact us for exact sizing.
| Component | Platform | Specifications | Qty | Role |
|---|---|---|---|---|
| AI-NVR Edge Node | Dual Hailo-8 (52 TOPS) on RDP carrier | 2× Hailo-8 M.2, 16GB RAM, 512GB NVMe, 8× PoE camera ports, 1U compact | 8–32 | 8 cameras per node — face detection, perimeter intrusion, crowd density, anomaly alert |
| IP Camera (Outdoor) | 4MP bullet / dome IP camera | IR 80m, WDR 140dB, H.265, PoE (802.3af), IP67, IK10, ONVIF-S | 64–256 | Perimeter, entry/exit, parking, open area coverage |
| IP Camera (Indoor) | 2MP dome IP camera | IR 30m, WDR 120dB, H.265, PoE, ONVIF-S, corridor format | As needed | Corridors, lobbies, server rooms, restricted areas |
| Central AI Hub | NVIDIA Jetson AGX Orin (275 TOPS) | 64GB RAM, 2TB NVMe, dual 10GbE, 2U rack, GPU-accelerated search | 1–2 (HA) | Alert aggregation, cross-camera tracking, forensic search, SOC dashboard |
| Storage Server | RDP NAS Server | 12-bay 3.5" HDD, 8× 16TB RAID-6, dual PSU, 10GbE, hot-swap | 2–4 | 30-day continuous recording at 4MP H.265 (128TB+ usable) |
| PoE Network Switch | Managed L2+ PoE switch | 24-port PoE+ (370W budget), 4× 10G SFP+ uplinks, VLAN, QoS | 4–12 | Camera power + data, network segmentation, QoS for video |
| SOC Workstation | RDP Desktop + Video Wall | Intel Core i7, 32GB RAM, multi-monitor (4–8 screens), VMS client | 1–4 | Security Operations Centre monitoring and alert triage |
What you need to deploy
Power
PoE for cameras (15W each), standard rack power for servers. Total: 5–15 kW depending on scale.
Cooling
Server room: 18–24°C, precision cooling. Edge nodes: fanless, 0–55°C rated. Cameras: outdoor-rated IP67.
Connectivity
Gigabit Ethernet backbone, 10GbE between racks, PoE to cameras. No internet required for core operation.
Space
Server room: 1–2 full racks. Edge nodes: distributed in IDF/comms rooms near camera clusters.
Installation
RDP provides site survey, camera placement design, cable routing plan, commissioning, and VMS configuration
Standards and regulatory alignment
- DPDP Act 2023 (data localisation)
- GDPR-ready architecture
- IS 16910 for CCTV systems
- CPWD / PWD specifications
- GeM-listed (all components)
- BIS certification
- STQC empanelment eligible
Ready to deploy Campus Surveillance Network?
Share your scale, site requirements, and timeline. We’ll customise this reference architecture to your exact deployment — from BOM to commissioning.