LIVE SYSTEM DEMO

Autonomous Border
Intelligence

Fixed sensor towers + autonomous drones + AI analysis. Every camera feed processed in real time. Every anomaly detected, classified, and actioned — before a human even looks at the screen.

CAM-04 · SECTOR 7-N · KUWAIT BORDER

03:47:22

VEHICLE · 94%

PERSON · 87%

⚠ THREAT DETECTED — DRONE DISPATCHED

CAM-11 · SECTOR 3-E · SAUDI BORDER

03:47:22

WILDLIFE · 91%

CLEAR — BASELINE NORMAL

DRONE-02 · RESPONDING · ALT 45M

03:47:31

TRUCK · UNLIT · 96%

ETA TARGET: 00:38

DRONE VISUAL — AWAITING ID

CAM-07 · MARITIME · UAE COAST

03:47:22

VESSEL · TAGGED · 99%

VESSEL · TAGGED · 97%

2 VESSELS — BOTH AIS CONFIRMED

CAM-UKR · FRONTLINE SENSOR · ZAPORIZHZHIA

05:47:22

VEHICLE MOVEMENT · 88%

ACTIVITY — CROSS-REF POI SYSTEM

CAM-SGP · STRAIT OF MALACCA · BEARING 270

09:47:22

TANKER · IMO MATCHED

PATROL · RSAF

ALL VESSELS IDENTIFIED — CLEAR

// LIVE

⚠ 03:47 — CAM-04 KWT: UNLIT VEHICLE + PERSON — DRONE DISPATCHED ✓ 03:44 — CAM-07 UAE: 2 VESSELS AIS CONFIRMED — CLEAR ◈ 03:41 — CAM-UKR ZAP: VEHICLE MOVEMENT — FLAGGED TO POI SYSTEM ✓ 03:38 — CAM-11 KSA: WILDLIFE CLASSIFIED — NO ACTION

// SYSTEM ARCHITECTURE

Why Flying Orbs Alone Don't Work — And What Does

Continuous flight requires continuous power. The right architecture uses fixed sensors as the brain and drones as the eyes-on-demand.

🗼

Fixed Sensor Towers

The permanent backbone — never sleeps

Placed every 8–15km along border or perimeter
PTZ cameras: optical (day) + thermal (night) + wide-angle overview
Radar for vehicle/person detection beyond camera range
Starlink backhaul — no ground cable needed
Solar + battery power — fully off-grid capable
AI runs locally on edge compute — works during comms outages

Coverage per tower: 2–8km radius · Uptime: 99.7% · Cost: $15K–$40K/tower

🚁

Autonomous Drone Response

Eyes-on-demand — launched by AI alert

Drones parked at automated charging stations every 20–30km
Auto-launched when tower AI flags anomaly needing closer look
Arrives at target within 2–4 minutes of alert
Transmits live HD video directly to command dashboard
Flies autonomously to GPS coordinate — operator takes over if needed
Returns to station, recharges, ready again in 20 minutes

Response time: <4 min · Flight time: 35–45 min · Recharge: 20 min

🧠

AI Intelligence Layer

What you build — runs on top of any hardware

Processes every camera feed 24/7 — no human watching screens
Classifies every object: person, vehicle type, animal, vessel
Behavioural analysis: approaching, stationary, digging, fleeing
Baseline learning: "normal" for each camera at each time of day
Anomaly scoring: how unusual is this vs the past 30 days?
Claude reasoning: generates structured alert with recommended action

Latency: <800ms detection · False positive rate: <3% after 30-day baseline

📟

Command Dashboard

One operator, entire border

Map view: all towers, all drones, all active alerts at once
Alert queue: prioritised by threat score — highest first
One-tap drone dispatch to any alert location
Live feed mosaic: jump to any camera in one click
Incident log: every detection timestamped, GPS-tagged, archived
Runs in any browser — tablet, laptop, command centre screen

Operators needed: 1–3 per shift vs 50+ for manual monitoring

// AI PROCESSING PIPELINE

What Happens in Under 800ms

From raw pixel to structured intelligence — fully automated, every frame, every camera.

01 · INGEST

Video Stream

RTSP stream from any camera — tower PTZ, drone gimbal, aerostat camera, existing CCTV. Standardised input layer accepts any source.

RTSP/WebRTC FFmpeg Edge buffer

02 · DETECT

Object Detection

YOLOv8 fine-tuned on military/border context. Detects: people, vehicles by type, vessels, aircraft, animals. Runs on edge GPU — no cloud needed.

YOLOv8 NVIDIA Jetson Custom weights

03 · TRACK

Movement Analysis

Multi-object tracking across frames. Direction, speed, trajectory prediction. Behaviour classification: approaching border, stationary loitering, retreat, digging.

ByteTrack Kalman filter Trajectory model

04 · SCORE

Anomaly Engine

Compares current detection against 30-day baseline for this camera, this time of day. Scores 0–100: how unusual is this? Filters out daily patterns automatically.

Baseline model Time-of-day weight PostgreSQL

05 · REASON

Claude Alert

High-score detections sent to Claude API. Claude generates: threat classification, recommended response, confidence, cross-reference with active intel (POI system, Swarm Brain).

Claude API Structured output Intel cross-ref

// SAMPLE OUTPUT

{
  "alert_id": "KWT-CAM04-20260415-0347",
  "threat_score": 87,
  "classification": "UNLIT VEHICLE + PERSON — BORDER APPROACH",
  "objects": ["truck_medium", "person_x1"],
  "behaviour": "approaching_border_northeast_4kmh",
  "anomaly_vs_baseline": "unlit vehicle at 03:47 — 0 prior occurrences in 30 days",
  "recommended_action": "DISPATCH DRONE — intercept at grid 2847-1193",
  "cross_reference": "POI system: no elevated threat on this axis today",
  "drone_dispatched": true, "eta_seconds": 142
}

// WHAT THE AI DETECTS

🚶

Person / Group

Confidence: 85–97% · Range: up to 800m (thermal)

Individual or group crossing detection. Head count. Carrying objects. Direction and speed. Distinguishes from wildlife via gait analysis.

🚛

Vehicle Classification

Confidence: 88–98% · Type: 12 classes

Car, truck, pickup, motorcycle, APC, tank, construction. Identifies if unlit (lights off at night = anomaly). Estimates occupant count.

⛵

Maritime Vessel

Confidence: 90–99% · AIS cross-ref

Vessel detection and type classification. Cross-references AIS transponder data. Flags vessels without AIS, unusual speed, or approach vectors.

⛏️

Digging / Construction

Confidence: 75–88%

Detects earthwork activity — IED emplacement, tunnel construction, fortification building. Compared against prior terrain baseline.

🦌

Wildlife Filter

Filter accuracy: 94%

Critically reduces false alarms. Distinguishes human vs animal movement by gait, size, thermal signature. Camel, goat, dog classified and filtered automatically.

🔥

Thermal Anomalies

Night detection range: 2–4km

Engine heat, body heat, fire, weapons discharge flash. Thermal cameras defeat darkness completely. No visible light required for detection.

// MARKET APPLICATIONS

Six Deployments. All from the Same Platform.

The AI layer is identical. The hardware configuration, threat models, and detection priorities are customised per country.

🇰🇼

Kuwait

Iraqi border — 254km land

HIGH PRIORITY

254km

Border length

~25

Towers needed

$4M–$8M

Deployment cost

Flat desert terrain — ideal for sensor towers. Minimal natural cover means long camera ranges. Key threat: smuggling, militia infiltration from Iraqi border regions. Currently monitored by US forces at Camp Buehring — Kuwait needs sovereign capability.

Cross-border smuggling Militia infiltration Vehicle border runs

🇸🇦

Saudi Arabia

Yemen border — 1,800km + Red Sea coast

CRITICAL

1,800km

Yemen border

~180

Towers needed

$25M–$60M

Full deployment

Active conflict zone — Houthi drone and missile launches, infiltration attempts, arms smuggling. Saudi already has physical fencing but zero AI analysis layer. The existing Elbit sensor network has no predictive intelligence integration. Largest addressable market in the region.

Houthi infiltration Drone launch sites Arms smuggling routes Ballistic missile prep

🇦🇪

UAE

Maritime coast + Strait of Hormuz

HIGH

1,318km

Coastline

Maritime

Primary threat axis

$10M–$25M

Phase 1

Maritime surveillance is the UAE's primary need — Strait of Hormuz passage, Iranian fast-boat threat, smuggling via sea routes. Tower network on coastal headlands with maritime radar integration. AIS cross-referencing flags vessels without transponders automatically.

Iranian fast boats Non-AIS vessels Maritime smuggling Underwater approach

🇶🇦

Qatar

Peninsula — maritime all sides

HIGH

563km

Coastline

Peninsula

Terrain type

$6M–$15M

Full perimeter

Qatar is a peninsula — maritime surveillance covers the entire perimeter. LNG infrastructure protection is the primary use case. North Field (world's largest gas field) requires perimeter monitoring. Also: Al Udeid Air Base perimeter AI for US forces.

LNG infrastructure Al Udeid perimeter Maritime intrusion

🇸🇬

Singapore

Strait of Malacca — maritime + port

STRATEGIC

193km

Coastline

~90K

Vessel transits/yr

$8M–$20M

Maritime layer

One of the world's busiest shipping lanes. Port security, maritime surveillance, and aerial approach monitoring. Singapore already has advanced radar systems — the AI intelligence layer on top of existing hardware is the gap. DSTA would fund this directly.

Piracy / maritime threat Port security Straits surveillance UAV intrusion

🇺🇦

Ukraine

Active frontline — 1,200km+

FRONTLINE

1,200km+

Active front

Feeds POI

System integration

$10M–$30M

Donor-funded

Frontline sensor network feeding directly into Swarm Brain and Predictive Offensive Intelligence. Visual detections of Russian vehicle movement cross-referenced with POI threat assessments. Every anomaly on a flagged axis gets elevated threat score automatically.

Armour assembly Infantry approach EW vehicle movement Artillery repositioning

// PRODUCT INTEGRATION

One Platform. Four Products. Compound Intelligence.

Border Intelligence feeds into every other product in the suite — each one making the others more valuable.

🔮

Predictive Offensive Intel

Visual detections of vehicle concentrations and logistics movement feed directly into the POI threat assessment engine. Ground truth confirmation of satellite-detected patterns.

→ Feeds: vehicle movement data, infrastructure changes

🎯

Swarm Brain

Confirmed vehicle positions from border sensors go straight into the drone mission target queue. No operator needed to log the target — the border AI does it automatically.

→ Feeds: confirmed target coordinates, vehicle type

🧠

Sovereign AI Analyst

Border detection data becomes part of the sovereign data platform. Ministers and commanders can query "what vehicle activity occurred on the northern border last week?" in natural language.

→ Feeds: incident log, pattern data, monthly summaries

🌏

Country Intelligence

Border activity patterns over time become geopolitical intelligence. Unusual troop movements, new infrastructure, civilian displacement — all automatically correlated and surfaced.

→ Feeds: long-term pattern library, threat evolution data

// HARDWARE — YOU DON'T MANUFACTURE

Off-the-Shelf Hardware. Your AI on Top.

You are not a hardware company. You build the intelligence layer that runs on cameras and drones the customer already has or can buy from established vendors.

PTZ Camera

Axis Q6135-LE

Long-range optical + IR. 32x optical zoom. RTSP stream. Industry standard for perimeter security. Your software connects via standard API.

~$3,000–$6,000/unit

Thermal Camera

FLIR Elara FR-345

Long-wave infrared. 3.5km detection range for vehicles. Fully integrates with standard video management systems via ONVIF protocol.

~$8,000–$15,000/unit

Surveillance Drone

DJI Dock 2 + M3T

Automated dock — drone auto-launches, lands, recharges without human. Thermal + optical payload. RTMP video stream. Available now, proven in field.

~$20,000–$35,000/dock

Edge Compute

NVIDIA Jetson Orin

Runs YOLOv8 detection at 30fps locally on tower. No cloud latency for detection. Sends only alert events to central server — minimal bandwidth.

~$800–$1,500/unit

Connectivity

Starlink Maritime

200Mbps uplink per tower. Works anywhere. No infrastructure required. Each tower is fully autonomous — power (solar) + comms (Starlink) + compute (Jetson).

~$2,500 hardware + $250/mo

Radar

Echodyne EchoFlight

Compact MESA radar. Detects people at 300m, vehicles at 1.5km. Triggers camera PTZ automatically to point at detected object. Works in total darkness, fog, dust.

~$15,000–$25,000/unit

// REVENUE MODEL

Hardware Pass-Through + AI Software License

You don't make money on cameras. You make money on the intelligence layer — priced as annual SaaS per km of border covered.

Pricing Model

Per-KM Software License

$8K–$15K

Per km of border per year. Includes AI platform, dashboard, updates, support. Hardware quoted separately at cost + 15% margin.

Example — Kuwait

254km Iraqi Border

$2M–$3.8M

Software ARR. Hardware one-off: $4M–$8M (25 towers + 8 drone docks). Total Year 1: $6M–$12M.

Example — Saudi Arabia

1,800km Yemen Border

$14M–$27M

Software ARR alone. Largest single contract in the product suite. Hardware: $25M–$60M one-off. Total Year 1: $40M–$90M.

Pilot

Free

10km stretch, 3 towers, 1 drone dock. 90-day proof. Customer provides hardware or you supply at cost.

First Contract

$2M–$5M

100km border segment. Software + implementation. Kuwait or Qatar as first customer.

Full Country

$6M–$28M ARR

Complete border deployment. Per-km SaaS recurring. Saudi Arabia alone is $14M–$27M/yr.

Suite Upsell

+40–60%

Every Border Intel customer is a natural buyer for Sovereign AI Analyst and Predictive Intel. Same trust, same infrastructure.

// THE INSIGHT

You're not selling cameras.
You're selling the intelligence layer that makes cameras
worth having in the first place.

Every country in this analysis already has some cameras, some towers, some sensors on their borders. None of them have AI that watches all of it simultaneously, learns what's normal, and tells operators exactly what to look at and why. That gap is worth $80K–$150K per km of border per year in software licensing alone — and it connects every other product in the suite.