"DroneRFz and DroneRFB: An Extension RF Spectrogram Dataset for Drone Recognition Across Diverse Electromagnetic Environments"

"The proliferation of unmanned aerial vehicles (UAVs) poses significant air-space security challenges, necessitating the development of robust drone recognition systems. Although radio frequency (RF)-based detection technologies have shown tremendous promise, existing public datasets remain scarce and lack category overlap among them, which severely limits the validation of models' generalization capabilities across varying real-world scenarios. To increase the diversity of drone categories and expand the scope of electromagnetic scenarios, we present an extension dataset comprising two novel subsets, DroneRFz and DroneRFB, specifically designed for drone recognition across complex and diverse electromagnetic environments. Acquired utilizing the HM X310 Software Defined Radio (SDR) platform, the dataset comprehensively captures RF signals across the 2.4 GHz and 5.8 GHz ISM bands. Specifically, DroneRFz encompasses signals from seven distinct drone models at distances up to 1 km, while DroneRFB focuses on the signal characteristics of two DJI models at a close range of 50 m. These two datasets were recorded in different urban environments, accompanied by severe spectral interference from Wi-Fi, Bluetooth, and impulsive bursts. Furthermore, we provide a rigorous data preprocessing pipeline that transforms raw I\/Q sequences into standardized 512 \u00d7 512 RF Time-Frequency Spectrograms (TFS), incorporating frequency center-cropping and Min-Max normalization to ensure cross-device feature consistency. The proposed dataset aims to provide strong support for the development, evaluation, and benchmarking of robust, deep learning-based UAV classification models."