Hybrid CLEAN filtering algorithm and terahertz radar for feature extraction of multi-rotors unmanned aerial vehicles

With the rapid advancement of unmanned aerial vehicle (UAV) technology and its wide application in modern warfare, UAV feature extraction has become increasingly prominent. However, during actual feature extraction processes, raw echo data are susceptible to interference from complex environmental noise, making it difficult to directly obtain feature information of UAVs. In this work, a hybrid CLEAN filtering algorithm and terahertz radar are introduced for feature extraction of the multi-rotor UAV. First, we use terahertz frequency modulated continuous wave radar to obtain scanning imaging of low altitude hovering multi-rotor UAVs. Subsequently, a bandpass filtering method is applied to determine target positions while dynamic filtering with an improved CLEAN algorithm is integrated to effectively eliminate noise. Finally, bilateral filtering is used to enhance edges and clearly delineate target contours. When conducting experiments with multi-rotor UAVs of DJI Mavic3 and DJI Mini3 at a distance of no more than 10 m, the height error is within 30 mm and the width error is less than 40 mm, respectively. The proposed method improves the identification capability of terahertz radar for "low, small, slow" targets.