Drone Detection & Classification with Surveillance ‘Radar On-The-Move’ and YOLO
Hani Haifawi (Robin Radar Systems, Student TU Delft)
Francesco Fioranelli (Microwave Sensing, Signals & Systems)
Alexander Yarovoy (Microwave Sensing, Signals & Systems)
Rob van der Meer (Robin Radar Systems)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
A new method to jointly detect and classify drones using a moving surveillance radar system (‘radar on-the-move’) and computer vision is presented. While most conventional counter-drone radar-based techniques focus on time-frequency distributions to obtain classification features, such approaches are limited in volumetric spatial coverage. To compensate for this, surveillance radars that offer full spatial coverage are used, but the determination of the best detection and classification approach to be applied on the resulting data is still an open challenge. In this paper a framework is proposed that combines deep learning approaches from computer vision, specifically the You Only Look Once (YOLO) network, with data from the moving surveillance radar produced by Robin Radar Systems B.V. This framework allows to jointly detect and label targets based on range-Doppler images generated in real-time. The method is validated on experimental data, with preliminary results on a small dataset showing precision, recall, mean average precision (mAP@0.5) and Area Under Curve (AUC) of over 99%