A new simulation methodology for generating accurate drone micro-Doppler with experimental validation

Abstract

Unmanned Aerial Vehicles, or drones, pose a significant threat to privacy and security. To understand and assess this threat, classification between different drone models and types is required. One way in which this has been demonstrated experimentally is through this use of micro-Doppler information from radars. Classifiers capable of exploiting differences in micro-Doppler spectra will require large amounts of data but obtaining such data experimentally is expensive and time consuming. The authors present the methodology and results of a drone micro-Doppler simulation framework which uses accurate 3D models of drone components to yield detailed and realistic synthetic micro-Doppler signatures. This is followed by the description of a purpose-built validation radar that has been developed specifically to gather high-fidelity experimental drone micro-Doppler data with which is used to validate the simulation. Detailed comparisons between the experimental and simulated micro-Doppler spectra from three models of drones with differently shaped propellers are given, showing very good agreement. The aim is to introduce the simulation methodology. Validation using single propeller micro-Doppler is provided, although the simulation can be extended to multiple propellers. The simulation framework offers the potential to generate large quantities of realistic drone micro-Doppler signatures for training classification algorithms.