Advancements in UAV Detection and Classification: Fusion of Mechanical Control and Sensor Data

AIML

This research investigates the classification of unmanned aerial vehicles (UAVs) using micro-Doppler signatures, a crucial technique for discerning between different types of UAVs. Through the utilization of a full-wave electromagnetic CAD tool, we explore the influence of control systems on the range–Doppler signatures of various UAV configurations, including quadcopters, hexacopters, and helicopters. Our approach integrates a mechanical control-based machine learning (ML) algorithm, with convolutional neural networks (CNNs) demonstrating robust performance, achieving classification accuracies exceeding 90%. Additionally, we introduce a novel methodology for simulating radar datasets using CAD tools, enabling the generation of diverse datasets tailored to different drone types and radar parameters. Leveraging a simulated 77 GHz frequency-modulated continuous wave (FMCW) radar, our classification model attains an accuracy of over 97%. We establish a theoretical framework linking micro-Doppler signatures with UAV motion dynamics, providing insights into spectral distribution. Experimental analysis, combining simulations and measured data, underscores the potential for effective detection and classification. Furthermore, we propose a ground-based surveillance radar system for drone detection, surpassing existing methods in both accuracy and computational efficiency. Validation through field experiments with a commercial portable radar underscores the viability and effectiveness of our approach. Overall, this study contributes to advancing UAV classification techniques and enhancing drone detection capabilities in practical scenarios.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.