Aerial Targets Classification Using Micro-Doppler Signatures

In recent years, the growing use of small unmanned aerial vehicles (UAVs) has raised significant national security concerns, necessitating advanced surveillance systems capable of identifying and classifying low radar cross-section (RCS) aerial targets. This paper proposes a transfer learning-based approach leveraging a deep convolutional neural network (DCNN) for the classification of low RCS aerial targets, employing the VGG16 model as a feature extractor to achieve high classification accuracy. Using a dataset comprising 4,849 micro-Doppler signature images, our approach demonstrates effective classification across five types of small aerial targets, including two-blade rotors, three-blade rotors, quadcopters, and bionic birds. In experimental validation, VGG16 attained a precision of 88%. false-negative rates. Overall, our results indicate the possibility of Developing efforts on aerial target classification and improving DCNN-based models, especially VGG16 for UAV detection They monitor activities for national security purposes.