Flying Object Tracker

The Flying Object Tracker Demonstrator explores the use of edge AI and radar to detect and classify drones and birds in real time. Growing use of unmanned aerial vehicles (UAVs) for delivery, inspection and recreation has led to safety concerns, while bird strikes threaten aircraft and wind turbines. Research by Hakani and Brahmat (2024) demonstrated that YOLOv9 running on an NVIDIA Jetson Nano achieved mean average precision (mAP) of 95.7 %, precision of 0.946 and recall of 0.864 for drone detection, with a 75× performance improvement over previous models. The system delivered real-time inference on edge hardware and underscored the feasibility of edge-based object detection. SmartLight’s demonstrator adapts these techniques to classify both drones and birds using a combination of mmWave radar and camera sensors. The radar provides range, speed and trajectory information; the camera supplies visual cues for classification. TinyML models run on a Jetson Orin and an ESP32 co-processor. The system processes data locally, triggering an alert when an unauthorised drone enters a restricted zone or when birds approach wind turbines or airports.