Research On Ionospheric Clutter Suppression Method For High Frequency Surface Wave Radar Based On Neural Networks

As an over-the-horizon detection radar,high-frequency surface wave radar adopts vertical polarization antenna,which realizes the detection of targets beyond the curvature of the earth by virtue of the characteristics of small attenuation of short-wave(3～30Mhz)propagation around the ocean.In the actual working process of radar,there are a variety of clutter interference,and ionospheric clutter often seriously affects the detection performance of radar,so this thesis will start from the characteristics of ionospheric clutter of highfrequency surface wave radar,combined with neural networks,establish a model of ionospheric clutter,so as to achieve the suppression of ionospheric clutter.The main work of this thesis is:Firstly,based on the measured radar echo data,combined with the structure of the ionosphere and the generation mechanism of ionospheric clutter,this thesis analyzes and verifies the distance-correlation,chaotic and time-frequency characteristics of ionospheric clutter.Then,according to the distance correlation and chaos characteristics of ionospheric clutter,the initial parameters are calculated,so as to construct a wavelet neural network,and the chaotic characteristics of ionospheric clutter are learned by training the neural network,and the classical particle swarm optimization algorithm is introduced and improved to improve the accuracy and stability of the neural network.Finally,the radar echo after ionospheric clutter suppression is obtained by making the difference between the predicted ionospheric clutter and the measured radar echo.Then,in order to solve the problem that the actual processing speed of ionospheric clutter suppression algorithm is too slow,combined with the time-frequency characteristics of ionospheric clutter,an ionospheric clutter suppression method based on time-frequency image processing is proposed.In the time-frequency domain,CNN-CGAN is used to extract and suppress ionospheric clutter images,and it is found that this method can suppress ionospheric clutter well,but it cannot be perfectly retained for the flooded target signal.Therefore,in order to solve the problem that the flooded target is not easy to retain in the time-frequency image processing model,a two-way long short-term memory neural network model is introduced,and an ionospheric clutter suppression method based on time-frequency image prediction is proposed,which is first amplified by CNN-CGAN data,and then CNNBiLSTM is used in the time-frequency domain Ionospheric clutter images are predicted,and ISTFT is finally used to achieve suppression in the time domain.After the experimental analysis of the actual measurement,the submerged target signal can be retained while suppressing ionospheric clutter.Experimental analysis of measured data simulation shows that this thesis has important application significance for realizing ionospheric clutter suppression and improving the detection accuracy of high-frequency surface wave radar on maritime targets.