Multi-sensor perimeter security for small-medium factories, campuses, and gated communities. Scalable edge AI solutions combining cameras with radar integration at vehicle entrances.
AI-powered surveillance for small-medium industrial facilities and gated communities. This guide covers deployments with 50–100 cameras and radar sensors at truck/car entrances.
50–100 IP cameras with centralized or distributed edge processing
Factory floor, perimeter fence, loading docks, vehicle entry/exit gates
Intrusion detection, LPR, PPE compliance, vehicle tracking, behavior analysis
2–4 units at truck/car entrances for vehicle detection and speed monitoring
Calculate power and network switch requirements before hardware selection.
| Component | Typical Power | ×50 Cams |
|---|---|---|
| PoE Camera (4K) | 12–25W each | 600–1,250W |
| PoE Camera (1080p) | 8–15W each | 400–750W |
| Edge Node (RK3588+Hailo) | 25W each | 50W (×2) |
| Edge Node (QCS8550) | 40W each | 40W (×1) |
| PoE Switch (24-port) | 30W + PoE budget | 60W (×2) |
| Radar Sensor | 10–20W each | 40W (×2) |
| Cameras | Switch Config | PoE Budget |
|---|---|---|
| 25 | 1× 24-port PoE+ | 370W |
| 50 | 2× 24-port PoE+ | 740W |
| 75 | 3× 24-port PoE+ or 2× 48-port | 1,110W |
| 100 | 2× 48-port PoE++ or 4× 24-port | 1,480W |
Formula: Cameras × Bitrate (Mbps) × 3600 × 24 × Days ÷ 8 ÷ 1,000,000 = TB
Example (50 cams, 4 Mbps, 30 days): 50 × 4 × 3600 × 24 × 30 ÷ 8 ÷ 1,000,000 = ~65 TB
With smart recording (motion only), reduce by 50–70% → ~20–30 TB
Three tiers optimized for different deployment requirements and budget constraints. All platforms support ONVIF cameras and standard VMS integration.
Complete architecture for a mid-range deployment: 50 cameras + 2 radars at vehicle entrances.
Typical small factory or gated community with single entry/exit point
| Component | Specification | Qty | Power | Network |
|---|---|---|---|---|
| IP Cameras (1080p) | 4MP turret/bullet, PoE, IR 30m, ONVIF | 40 | 12W each = 480W | 100Mbps each |
| IP Cameras (4K LPR) | 8MP bullet, motorized zoom, WDR | 10 | 18W each = 180W | 100Mbps each |
| Edge Node | RK3588 + Hailo-8 M.2 (32 TOPS) | 2 | 25W each = 50W | 2× GbE each |
| PoE Switch | 24-port PoE+ (370W budget), 2× SFP+ | 2 | 30W each = 60W | 10GbE uplink |
| Core Switch | L3 managed, 24-port GbE + 4× SFP+ | 1 | 35W | 10GbE backbone |
| VMS Server | Intel N200 Mini PC, 32GB RAM, 4TB NVMe | 1 | 25W | 2.5GbE |
| NAS Storage | 4-bay NAS, 4× 8TB HDD (32TB raw) | 1 | 50W | 2× GbE |
| Radar Sensor | 77GHz digital imaging, 150m range | 2 | 15W each = 30W | Ethernet / CAN |
| UPS | 1500VA / 900W, 30min runtime | 1 | — | — |
| Component | Power (W) |
|---|---|
| 40× 1080p Cameras (PoE) | 480W |
| 10× 4K LPR Cameras (PoE) | 180W |
| 2× Edge Nodes | 50W |
| 2× PoE Switches (base) | 60W |
| 1× Core Switch | 35W |
| 1× VMS Server | 25W |
| 1× NAS Storage | 50W |
| 2× Radar Sensors | 30W |
| Subtotal | 910W |
| + 20% Headroom | 182W |
| TOTAL POWER | ~1,100W |
| Segment | Bandwidth | Notes |
|---|---|---|
| Camera → PoE Switch | 50× 4Mbps = 200 Mbps | Per switch: 100 Mbps |
| PoE Switch → Edge Node | 2× 100 Mbps | Processed video streams |
| Edge Node → VMS | 2× 50 Mbps | Metadata + alerts + selective video |
| VMS → NAS | 200 Mbps sustained | Recording streams |
| Core Uplinks | 10GbE recommended | Future-proof, burst capacity |
| Item | Qty | Unit Price | Total |
|---|---|---|---|
| 4MP PoE Turret Camera | 40 | $80 | $3,200 |
| 8MP LPR Bullet Camera | 10 | $200 | $2,000 |
| RK3588 + Hailo-8 Edge Box | 2 | $1,000 | $2,000 |
| 24-Port PoE+ Switch (370W) | 2 | $350 | $700 |
| L3 Core Switch (24-port) | 1 | $500 | $500 |
| Intel N200 Mini PC (VMS) | 1 | $400 | $400 |
| 4-Bay NAS + 32TB HDD | 1 | $1,200 | $1,200 |
| 77GHz Digital Imaging Radar | 2 | $600 | $1,200 |
| 1500VA UPS | 1 | $300 | $300 |
| Cabling + Accessories | 1 | $500 | $500 |
Distributed Processing: Edge nodes handle AI inference locally, reducing central server load by 90%+.
Redundancy: Each edge node covers half the cameras; if one fails, the other continues.
Scalability: Add more edge nodes or upgrade to QCS8550 for 100+ cameras.
Radar Integration: Edge nodes receive radar data via Ethernet, fuse with camera feeds for vehicle tracking.
Complement camera coverage with radar sensors at vehicle entrances for all-weather detection and velocity measurement. Essential for truck/car gate monitoring.
Type: 4D Digital Imaging Radar
Range: 150–350m
Resolution: <1° angular
Interface: Ethernet / CAN
Use Case: Large vehicle tracking, classification
Power: 15–25W
Price Range: $1,500–2,500
Type: 77GHz FMCW Radar
Range: 100–200m
Resolution: 1.5° angular
Interface: CAN / RS-485 / Ethernet
Use Case: Traffic monitoring, industrial
Power: 10–15W
Price Range: $800–1,500
Examples: Steradian, waveant, local OEMs
Type: 77/79GHz mmWave
Range: 50–150m
Interface: UART / SPI / CAN
Use Case: Basic vehicle detection
Power: 5–10W
Price Range: $300–800
For true sensor fusion (combined tracking), use Jetson Orin with ROS2 or NVIDIA Isaac ROS. Entry/mid-tier platforms (RK3588, QCS8550) work best with radar as a separate alert trigger rather than fused tracking — the edge node receives radar events via MQTT/Ethernet and correlates with camera zones.
Recommended software components for edge analytics, VMS, and alerting.
| Software | RK3588 | QCS8550 | Jetson Orin | Notes |
|---|---|---|---|---|
| NVIDIA DeepStream | ❌ | ❌ | ✅ | Jetson only |
| Hailo Tappas | ✅ | ❌ | ✅ | Requires Hailo M.2 |
| RKNN-Toolkit2 | ✅ | ❌ | ❌ | Rockchip only |
| Frigate NVR | ✅ | ✅ | ✅ | All platforms |
| Nx Witness | ✅ | ✅ | ✅ | ARM supported |
Hardware requirements for small-medium industrial facilities and gated communities.
| Deployment | Cameras | Radars | Edge Nodes | PoE Switches | Power (Total) | Storage (30d) | Est. Budget |
|---|---|---|---|---|---|---|---|
| Small Gate Community | 25–30 | 1 | 1× RK3588+Hailo | 1× 24-port | ~500W | 15–20 TB | $5,000–7,000 |
| Medium Factory | 50 | 2 | 2× RK3588+Hailo | 2× 24-port | ~1,100W | 30–40 TB | $12,000–15,000 |
| Large Campus | 75–100 | 2–4 | 2× QCS8550 or 3× RK3588 | 2× 48-port or 4× 24-port | ~1,800W | 50–80 TB | $20,000–30,000 |
Standard Recording: Cameras × 4 Mbps × 0.04 × Days = TB
Smart Recording (motion): Reduce by 50–70%
Example: 50 cams × 4 Mbps × 0.04 × 30 days = ~24 TB (smart: ~8–12 TB)
Edge nodes process AI inference locally and trigger real-time alerts directly — no cloud dependency for critical notifications.
| Method | Latency | Complexity | Snapshot | Best For |
|---|---|---|---|---|
| MQTT | <100ms | Low | Via URL | IoT dashboards, Node-RED, Home Assistant |
| Webhook/REST | <500ms | Low | Base64/URL | VMS integration, custom backends |
| Telegram Bot | <1s | Low | ✅ Inline | Security guards, mobile alerts |
| WhatsApp Business | <2s | Medium | ✅ Inline | Management alerts |
| Email (SMTP) | 1-5s | Low | ✅ Attachment | Logs, non-urgent alerts |
| Siren/Relay (GPIO) | <50ms | Low | N/A | Physical alarms, strobe lights |
Edge Node detects event → publishes to MQTT → Telegram Bot sends photo + alert to security group → if no acknowledgment in 2 min → escalate to WhatsApp/Email (management)
VikaBot provides end-to-end industrial surveillance solutions — from hardware sourcing to system integration and ongoing support.
Request Consultation