Simulating UAV micro-Doppler using dynamic point clouds

Abstract

The ability to classify between different unmanned aerial vehicle (UAV) types is a pressing security requirement in order to assess the capabilities and resulting threat of a UAV. It has been demonstrated experimentally that different UAVs can be classified using their micro-Doppler signatures. This classification exploits differences in the shapes of UAV components, particularly the propeller blades. Effective classification using micro-Doppler will require large amounts of training data but obtaining this experimentally is logistically difficult and potentially expensive. This paper presents a simulation approach that models UAV components as point clouds derived from accurate 3D CAD models. Simulated results at millimeter wave frequencies for the DJI Phantom 3, DJI S900 and Joyance JT5L-404 propellers are presented and compared with experimental results.