A real-time drone and bird detection system that uses synchronized cameras and stereo vision to reconstruct aerial objects in 3D space — then classifies them by flight behavior, not just appearance.
See Everything. Understand What Matters.
Modern airspace threats don't announce themselves. Drones are cheap, small, and fast — and by the time conventional systems flag them, it's often too late. Thunder Eye was built to close that gap: a compact, intelligent detection platform that turns raw video into real aerial intelligence, in real time.
Beyond Motion Detection
Most detection systems ask one question: is something moving? Thunder Eye asks a better one: what is it, where exactly is it, and what is it doing?
Using an array of synchronized cameras and stereo vision, Thunder Eye detects moving aerial objects and reconstructs their position in three-dimensional space. The result isn't a blip on a 2D screen — it's a live, continuous 3D model of low-altitude airspace above your site.
How It Works
1. Stereo Vision & 3D Reconstruction Synchronized cameras capture overlapping fields of view. Graph-based triangulation reconstructs each detected object's precise position in 3D space — giving operators altitude, bearing, and distance, not just a pixel coordinate.
2. Persistent Object Tracking A high-performance C++ processing pipeline tracks each object across frames with low latency, maintaining identity even as targets move, accelerate, or briefly occlude. Objects are followed continuously, not re-detected from scratch each frame.
3. Behavior-Based Classification This is where Thunder Eye diverges from conventional systems. Rather than relying solely on visual appearance — which can be ambiguous at distance — Thunder Eye analyzes trajectory data over time:
- Speed and acceleration profiles
- Path consistency and turning patterns
- Hovering behavior and altitude variation
- Flight envelope characteristics
From this, the system classifies objects into meaningful categories: drones, birds, aircraft, insects, and other airborne motion. Two objects that look identical in a single frame can be reliably separated by how they fly.
Built for the Edge
Thunder Eye's processing pipeline is written in C++ and engineered for efficient operation on edge hardware — no cloud dependency, no latency introduced by a round-trip to a remote server. Detection, tracking, and classification happen locally, where the cameras are deployed.
This makes the system:
- Passive & stealthy — no active emissions; undetectable by enemy electronic warfare
- GPS-independent — fully operational in contested or GPS-denied environments
- Low-latency — actionable output within milliseconds of detection
- C2-ready — output connects to military GIS and command-and-control dashboards
Technologies
- Computer Vision & Stereo Imaging
- Graph-Based 3D Triangulation
- Trajectory Analysis & Behavioral Classification
- High-Performance C++ Edge Processing
- AI & Machine Learning
- Defense Systems Integration & Optical Intelligence
Partners
Thunder Eye is developed in partnership with the Lithuanian Riflemen's Union, providing real-world operational context and field validation for counter-UAS scenarios.
Use Cases
- Perimeter monitoring — persistent low-altitude awareness around critical sites
- Infrastructure protection — early warning for energy, transport, and communications facilities
- Counter-UAS operations — detection and classification of loitering munitions and swarm threats such as Shahed/Geran-type drones
- Broader airspace observation — scalable deployment across wide geographic areas
Status
Active — Project initiated May 2025. Developed in collaboration with the Lithuanian Riflemen's Union, with field testing underway in counter-UAS configurations.
Interested in deploying Thunder Eye for your facility or operation? Contact us — we respond within 24 hours and are NDA-ready.