⚠ Disclaimer: This entry may be incomplete, out of date, or inaccurate. It is AI-maintained on a best-effort basis. Do not rely on it as a sole source — verify claims independently using the sources listed below.
Summary
Electro-optical (EO) and infrared (IR/thermal) camera systems detect drones visually against the sky or by heat signature from motors. EO/IR is the only passive sensor modality effective against fiber-optic tethered drones — which emit no RF, produce minimal acoustic signature at distance, but are still physically visible. Machine learning object detection (YOLO-class models) enables automated classification, but drone-vs-bird discrimination at range requires high-resolution sensors and well-tuned models. The Marduk Technologies Piraya system, tested at Siil 2025, is the current state-of-the-art for passive optical C-UAS.
Key Facts
- Type: Passive sensor (EO cameras require visible light; IR is passive; no emissions)
- Effective range: 200 m–2+ km depending on sensor resolution, aperture, and drone size
- Bird discrimination: Moderate to good with trained AI; challenging for small drones vs. small birds at range
- Works against: All drone types including fiber-optic and RF-dark; also detects birds (which then need to be filtered)
- Key advantage: Only modality that detects fiber-optic tethered drones at tactically useful range
- Key limitation: Requires adequate visibility (degraded by fog, heavy rain, darkness without IR); computational cost of real-time AI inference across multiple camera fields
What It Is / How It Works
Visible (EO) detection: High-resolution PTZ cameras or fixed wide-angle cameras cover a sector of sky. AI object detection models (YOLO, SSD, or custom architectures) run inference on frames to detect small moving objects. The primary challenge is distinguishing small drones from birds at range — both appear as small, dark, moving objects against sky. Detection range for a DJI Phantom-class drone is approximately 200–800 m with commercial megapixel cameras; purpose-built systems with larger apertures extend to 1–2 km.
Infrared (IR/thermal) detection: Drone motors and ESCs generate heat (typically 40–80°C during operation). MWIR and LWIR cameras detect this thermal signature against the cooler sky background. IR detection is less affected by lighting conditions (effective day and night) but thermal contrast varies with ambient temperature and is lower in hot weather. IR is particularly effective for detecting hovering drones where the motor has been running.
AI classification for drone-vs-bird:
- Visual object detectors (YOLO-class) train on annotated drone datasets. Public datasets include “Drone vs. Bird” detection challenge datasets (used in academic competitions).
- Motion analysis — trajectory prediction and flight path smoothness — helps filter out birds, which exhibit more erratic movement than drones on programmed paths.
- Multi-frame tracking improves accuracy over single-frame detection.
Fiber-Optic Drone Specific Application
Fiber-optic tethered drones are emerging as a critical threat. These drones:
- Carry no RF transceiver
- Are unjammable and undetectable by RF sensors
- Are being deployed in conflict zones (Ukraine/Russia) for both surveillance and kamikaze missions
- Are becoming commercially available
Why optical is the solution: While fiber-optic drones hide from all RF-based sensors, they cannot hide from optical observation — they are physically present in the sky and produce heat from motors. Marduk Technologies’ approach uses passive high-resolution optical and thermal cameras with AI tracking, specifically targeting this threat class.
NATO urgent challenge (April 2025): NATO issued an innovation challenge seeking detection solutions for fiber-optic FPV drones, with required detection range of 300–500 m — indicating this is considered an acute operational gap.
Key Commercial Implementation: Marduk Piraya
Marduk Technologies (Tallinn, Estonia; founded 2016) has developed the Piraya passive optical C-UAS system. Key facts from Siil 2025 exercise:
- Integrated on CV90 infantry fighting vehicle
- Six quadcopters detected, tracked, and neutralized
- One engagement at 2 km range
- Fully passive — emits no signals; operates in electronically contested environments
- Estonian Ministry of Defense funding support
Drone-vs-Bird in Visual Systems
Visual drone-vs-bird discrimination remains an active research area:
- Shape analysis: Multi-rotor drones have symmetric, compact shapes; birds have elongated bodies with outstretched wings — detectable at sufficient resolution
- Flight behavior: Drones maintain steadier flight paths; birds bank and flap irregularly
- Motor glint: Exposed ESCs and motors can produce visual/IR glints distinguishing them from birds
- The “Improving Small Drone Detection Through Multi-Scale Processing and Data Augmentation” research (drone-vs-bird detection challenge) represents current state of the art in academic benchmarking