Aircraft Target Classification Method Based On Narrow-band Radar Data

Grounded on radar target detection,radar target classification and recognition is a technology to classify and identify target attributes,categories or types based on electromagnetic scattering mechanism,radar echo signals of target and environment and target features extraction.This technology is widely used in military and civilian fields such as identification of friend or foe,identification of true and false warheads,detection of faulty targets,and exploration of petroleum deposits.Narrowband radar is the most widely used radar because of its mature technology,simple structure,strong anti-jamming capability and low price,so it is of great significance to maximize the ability of narrowband radar to identify targets.Therefore,based on narrowband radar data,this thesis conducts research on aircraft target size classification,aircraft target type(helicopter / propeller / jet)classification,and helicopter rotor physical parameter extraction.The main contents are as follows:(1)A method of aircraft target size classification based on non-uniform quantized state transition features is proposed.This method uses non-uniform quantization to re-divide the data distribution of the target,and determines the target size through the gap comparison of the state transition matrix.Compared with the classification method based on traditional statistical features(mean,variance,etc.),this method uses the fluctuation information of the target RCS more carefully,and has low angular sensitivity to RCS.Simulation experiments show that the classification effect using non-uniform quantized state transition features is more stable than the traditional statistical features,and the overall classification accuracy is slightly higher.Under the signal-to-noise ratio of 26 dB,the classification accuracy of aircraft target type of this method can reach more than 90%.(2)An aircraft target classification method based on the periodic characteristics of the first eigenmode function is proposed.This method uses the EMD algorithm to eliminate the fuselage components and decompose the micro-Doppler signal.The cyclic autocorrelation function is used to reduce the influence of noise and strengthen the periodic characteristics.Finally,the periodic characteristics are used to classify the target type of the aircraft.This method extracts micro-Doppler information from the radar echo of the target as a whole,which is more closed to the actual situation than the classification of only the ideal aircraft target rotor signal discussed in most papers.Simulation experiments verify the feasibility and effectiveness of the method.With a signal-to-noise ratio of 26 dB,the accuracy of the method for aircraft target types can reach more than 90%.(3)A method for helicopter physical parameter extraction and recognition based on time-spectrum image processing is proposed.This method extracts the time-frequency signal line from the time-frequency spectrum of the helicopter radar echo,and uses the least squares pairing method to fit the time-frequency signal line so that the physical parameters of the target can be estimated.Through these parameters,the model of the target helicopter can be identified easily.Compared with the deep learning-based target recognition method,this method only needs a small amount of training data to determine the threshold.Simulation experiments show that this method can accurately estimate the number of blades of the helicopter rotor,and the estimation error of the rotor speed and blade length is less than 10% in most cases.And the accuracy of helicopter model recognition can reach about 85% by using this method.