Convolutional Neural Network Assisted Detection and Localization of UAVs with a Narrowband Multi-site Radar

We present an approach to detect and locate non-cooperative UAVs from their micro-Doppler signature using a narrowband radar in a multi-site configuration. We describe a method for the localization of rotating objects with the geometric information obtained exclusively from their micro-Doppler signatures. This approach only requires very simple transceivers with CW waveforms, in a cost-effective multi-site architecture. A convolutional neural network is used to detect and identify the UAVs by extracting the characteristic features of their micro-Doppler signature. We present simulated and preliminary experimental data that show the technical viability of this concept.