PyroSentinel is an independent research and development project building a drone-based wildfire early detection system — integrating LWIR thermal imaging, onboard inference, and real-time ground alerting.
▶ Contact UsA multi-rotor UAV platform equipped with a stabilised LWIR thermal camera for continuous wide-area patrol. Designed for fully autonomous waypoint missions over designated high-risk forest zones.
LWIR thermal imagery is processed onboard to identify anomalous heat signatures indicative of early-stage ignition — before visible smoke or flame is detectable by conventional sensors.
GPS-tagged detection events are transmitted in real-time to a ground station, enabling rapid human-in-the-loop verification and escalation to fire response services.
Wildfire frequency and severity have increased markedly across East Asia over the past decade. In South Korea, the spring of 2025 saw the largest wildfire outbreak in recorded national history, with multiple simultaneous fires across Gyeongbuk province burning for several days and displacing thousands of residents.
The critical window for suppression is the first 10–20 minutes after ignition. Conventional detection methods — ground-based observation towers, manned patrol aircraft, and satellite imagery — are constrained by terrain, visibility conditions, and revisit intervals that frequently exceed this window.
PyroSentinel addresses this gap by deploying low-cost autonomous UAVs capable of continuous thermal surveillance over at-risk zones. By detecting heat anomalies before visible flame develops, the system aims to compress detection-to-alert time to under five minutes — enabling faster mobilisation of ground response teams.
The project is currently in the component evaluation and prototype development phase, with field testing planned for late 2025.
Uncooled LWIR microbolometer technology detects thermal radiation in the 8–14 μm band, enabling passive detection without illumination — effective in total darkness and through light smoke, where RGB cameras fail.
Pre-programmed waypoint missions provide systematic coverage over designated high-risk areas. Mission logic accounts for sensor recalibration cycles, ensuring continuous valid imagery throughout patrol duration.
A lightweight detection model runs on edge compute hardware colocated with the imaging system. Onboard inference reduces dependency on uplink bandwidth and enables faster event response without round-trip latency.
Detection events are packaged with GPS coordinates, timestamp, and thermal thumbnail and transmitted to a ground station over a dedicated RF link. An operator interface presents alerts in real time for human verification.
We are currently in the component procurement and evaluation phase. We welcome enquiries from thermal sensor manufacturers, module suppliers, and technology partners interested in supporting or collaborating on the project.
planiitske@gmail.com