Fuhrmann, L.L.FuhrmannBiallawons, O.O.BiallawonsKlare, J.J.KlarePanhuber, R.R.PanhuberKlenke, R.R.KlenkeEnder, J.J.Ender2022-03-142022-03-142017https://publica.fraunhofer.de/handle/publica/40189810.23919/IRS.2017.8008142Recent critical instances have demonstrated the demand for an effective way of detecting and classifying small Unmanned Aerial Vehicles (UAVs) as they pose a serious threat in civil security. We present results of radar measurements with a one channel continuous wave system at Ka band aiming at classifying UAVs through a detailed micro-Doppler analysis. High-sensitivity measurements of different UAVs (quadcopters of different size, octocopter, helicopter, fixed-wing plane) with a large number of different trajectories and flight parameters were obtained. Our analysis is based on different time-frequency transforms (Short-Time Fourier Transform, Cadence Velocity, Cepstrogram), followed by different feature extraction methods including a singular value decomposition. We present first classification results based on a Support Vector Machine algorithm for two different cases: (i) a global classification of the measured UAVs as man-made objects against a set of simulated flying bird data, and (ii) classification and characterization of different types of UAVs. In the latter case we also extract parameters such as number of rotors, rotation rate and rotor blade length. Our first results indicate very good classification accuracies ranging between 96% and 100 %.en621conference paper