Enhanced UAV Detection and Classification With Birds Using NLFM Pulse-Doppler Radar

Detecting UAVs in clutter environments and classification with bird is a difficult and important problem. The high sidelobe levels of typical linear frequcny modulation waveform limit the ability to detect UAVs without applying windowing methods with significant SNR reduction. In this paper, nonlinear frequency modulation (NLFM) waveforms are proposed to enhance the detection probability of UAVs. The comparisons between LFM and NLFM show that NLFM provides the improved performance in UAV signal detection. Furthermore, a simple clutter classification algorithm is proposed to classify microdoppler containing signals, such as UAVs or birds, from clutter. The NLFM waveform has low clutter side lobes and produces high power spectral density in the main lobe, which preserves and enhances microdoppler capabilities more effectively than those of LFM. Taking advantage of these features, a convolutional neural network (CNN) based classification of UAVs and birds using the doppler spectrum is demonstrated. The performance of the NLFM waveform was validated using the proposed lightweight CNN model. The doppler spectrum features derived from NLFM have achieved the improved detection and high performance.