Deep Learning for Classification of Mini-UAVs Using Micro-Doppler Spectrograms in Cognitive Radar

A Feature-Based Approach for Loaded/Unloaded Drones Classification Exploiting micro-Doppler Signatures

“…Approaches investigated features extracted from the received I/Q radar signal using empirical mode decomposition (EMD) [1] as well as extraction of features from both high resolution range profiles and Doppler [2]. The use of deep learning for the challenge of drones classification is also being investigated with approaches such as the one presented in [3] that demonstrated how multi-layers perceptron can provide high classification rates of drones with different propellers, while the method proposed in [4] has shown how deep learning can be useful to denoise the micro-Doppler signature before a classification stage. The work developed in [5], instead, has investigated how the micro-Doppler signature is affected by different weights of the payload and proposes additional strategies to classify the radar returns in a multi-static radar system.…”

A three-step classification framework to handle complex data distribution for radar UAV detection

“…UAVs using radars [6,11,15,[17][18][19][20][21][22]28]. Artificial neural networks were applied on spectrum directly to classify different types of UAVs [28].…”

“…Huizing et al employed Alexnet and LSTM-RNN on spectrograms to classify mini-UAVs [11], whereas Kim et al utilized GoogLeNet on the image merged from spectrogram and CVD [20]. Similarly, SVM was applied on the feature vector obtained from spectrogram and CVD [21].…”

“…Early techniques using kinematic and radar cross-section characteristics [3] could not reliably differentiate UAVs from birds, as the both kinds are small and slow-moving targets. Many representations of micro-Doppler signature (mDS) have been explored, e.g., spectrogram [4][5][6][7][8][9][10][11], cepstrogram [12], cadence velocity diagram [13][14][15], others [16][17][18][19], and combinations of the aforementioned [20][21][22]. Particularly in [22], a rich source of features including spectrogram, cepstrogram and CVD are utilized.…”

A Comparison of the Recursive and FFT-based Reassignment Methods in Micro-Doppler Analysis

“…In recent years, micro-Doppler analysis has become one of the fundamental techniques in target recognition and classification [1], [2], [3], [4], [5]. This is the result of a relatively simple analysis of the narrowband signal in the baseband, which allows fast algorithms to be applied.…”