DroneRFb-Spectra: A RF spectrogram dataset for drone recognition

To develop radio frequency-based drone recognition, we release an RF spectrogram dataset, named DroneRFb-Spectra. All signals of drones were collected by a Universal Software Radio Peripheral (USRP) device, recording three Industrial Scientific Medical (ISM) bands under urban scenarios. The classes cover 7 common brands, i.e., DJI, Vbar, FrSky, Futaba, Taranis, RadioLink, and Skydroid, with a total number of 14460. Each spectrogram has been downsampled to the size of 512x512 from the original IQ data with a length of 50ms by using the short-time Fourier transform.

为开展基于射频的无人机识别研究，我们公开发布了一款名为DroneRFb-Spectra的射频频谱图数据集。所有无人机信号均通过通用软件无线电外设（Universal Software Radio Peripheral，USRP）采集，于城市场景下记录了3个工业、科学、医疗（Industrial Scientific Medical，ISM）频段的信号。数据集涵盖DJI、Vbar、FrSky、Futaba、Taranis、RadioLink及Skydroid共7个主流品牌，总样本量达14460。所有频谱图均通过短时傅里叶变换（short-time Fourier transform），将时长为50ms的原始IQ数据转换并下采样至512×512的尺寸。

搜集汇总

背景与挑战

背景概述

DroneRFb-Spectra是一个用于无人机识别的射频频谱图数据集，包含通过USRP设备在ISM频段收集的14460个样本，覆盖7个常见品牌。每个频谱图通过短时傅里叶变换处理为512x512大小，旨在支持基于信号处理和深度学习的无人机识别方法研究。