Micro-Doppler spectrum correlation matrix feature extraction method of multi-rotor unmanned aerial vehicle

Abstract

The invention belongs to the technical field of multi-rotor unmanned aerial vehicle classification, and particularly relates to a micro Doppler spectrum correlation matrix feature extraction method ofa multi-rotor unmanned aerial vehicle. According to the method, firstly, short-time Fourier transform (STFT) is carried out on echo data of rotors of the multi-rotor unmanned aerial vehicle to obtainmicro-Doppler spectrum data, and then correlation matrix characteristics are extracted by utilizing the micro-Doppler spectrum data to realize identification of the multi-rotor unmanned aerial vehicle. Due to the fact that the extracted correlation matrix features only contain the micro Doppler information of the strong scattering points on the rotating blade and noise intensity is reduced, compared with physical parameter features extracted from time-frequency spectrum data, the method has better robustness to noise, and a higher recognition rate can be obtained even under the condition of alow signal-to-noise ratio.

Description

technical field [0001] The invention belongs to the technical field of classification of multi-rotor UAVs, and in particular relates to a micro-Doppler spectrum correlation matrix feature extraction method of multi-rotor UAVs. Background technique [0002] UAV is a special type of aerial target. With the progress of the military era, it is necessary to quickly classify which type they are, whether they are reconnaissance aircraft, decoys, electronic countermeasures, communication relays, target drones, or unmanned fighter jets. , which is of great significance to strategic adjustment. Since UAVs are unmanned and their physical size is relatively small, the classification rate obtained by classifying them by high-resolution radar range profile (HRRP) is not high at present, because sub-centimeter resolution is required to capture objects smaller than 100cm target space structure, and when the target line-of-sight angle has a small change, the one-dimensional range image will...

AI Technical Summary

Problems solved by technology

[0004] At present, the classification methods for UAVs mainly include two categories: the first method mainly uses the micro-Doppler spectrum of the UAV as a two-dimensional image, and uses the image recognition method to complete the recognition of the UAV, but this kind of method In the case of a low signal-to-noise ratio, the correct recognition rate drops significantly; the second type of method mainly extracts the physical parameters of the moving parts (such as the number of blades, the length of the rotor, the rotational speed, etc.) This type of method is still susceptible to the influence of noise. Under a low signal-to-noise ratio, accurate feature extraction cannot be performed, resulting in reduced recognition performance.

Images