Sea Clutter Doppler Characteristics Extraction And Prediction Based On Deep Learning

Sea clutter is referred to as scattering radar echo from the sea surface,and sea clutter Doppler characteristics is the very important part among all the characteristics of sea clutter.The Doppler characteristics of the sea clutter refers to the Fourier transformation of autocorrelation function of continuous coherent time-series of sea clutter signals,and related spectral parameters include Doppler shift and Doppler width.With the steady development of modern computation nature,machine learning methods reveal comparative advantages when processing large amount of data.However,deep learning methods applied in radar sea clutter remain sparse,and intellectual research on sea clutter Doppler characteristics is still just beginning.In this paper,shore-based radar real data is used for deep learning-based sea clutter Doppler parameters extraction and prediction.The specific contents and creative points in the paper are listed below.1.The background information about sea clutter Doppler characteristics is introduced.Besides,the significance of the research,as well as major achievements at home and abroad about intellectual application of sea clutter Doppler characteristics are elaborately demonstrated.Also,the background knowledge of sea clutter Doppler characteristics modeling and Doppler parameters estimation,as well as contents about sea wave spectrum are introduced,so as to offer theoretical bedding for the latter chapters.2.The preprocessing procedure of sea clutter real data is introduced,and a BackPropagation Neural Network-based Doppler parameters extraction method is raised.Based on the new method,the Doppler frequency and Doppler width are extracted from the shore-based radar real data,and the results of sea clutter Doppler parameters extraction under different numbers of pulses are analyzed.3.A Deep Neural Network-based Sea clutter Doppler parameters prediction method is proposed.Based on the parameters of radar measurement and marine environment,the Doppler parameters of the radar sea clutter are predicted.Furthermore,the prediction results are compared with those of various classical regression-prediction models.4.Sophisticated analysis is performed on the influence of the input parameters of the model and prediction results.Based on sea wave spectral models and the results in sea clutter Doppler parameters prediction,the input parameters are further optimized.Finally,based on the weight-transmitting theory of deep neural network,the prediction model of the Doppler parameters is explicitly fitted.