Augmentation of Training Data for AI-based Drone Detection System

Abstract

The increasing use of FPV drones in modern conflicts necessitates compact and energy-efficient detection systems capable of operating in dynamic electromagnetic environments. AI-based RF signal detection is a promising solution. However, its application can be limited by the lack of labelled datasets and the constraints of embedded platforms. This article presents a method for generating and augmenting training data directly from signals captured from analogue FPV video transmitters. Finally, a convolutional neural network was trained using the generated dataset and evaluated in a real-time environment. Experimental results demonstrate reliable detection performance, indicating that the proposed method is an effective and efficient solution for embedded FPV drone detection systems.