High Frequency Surface Wave Radar Clutter Database Establishment And Clutter Classification Algorithm Research

High-Frequance surface wave radar(HFSWR)detects targets beyond visual range by using the principle that high frequency electromagnetic waves can propagate around the coastal surface,which can effectively complement China's coastal defense capability and strengthen the surveillance and protection capability of territorial sea resources.However,the radar echo is faced with a variety of clutter and interference,which has a great impact on target detection.Nowadays,the widely used clutter suppression algorithm mainly deals with a single type of clutter to reduce the impact of this kind of clutter on target detection,but it has high limitations.If the clutter can be classified and identified before using the clutter suppression algorithm,the performance of the clutter suppression algorithm will be greatly improved,and the radar intelligent processing will be realized.In order to classify and recognize different types of clutter in HFSWR,firstly we need to analyze the correlation characteristics of clutter and establish the clutter feature database: original channel data features,amplitude features after beamforming,range features,Doppler features,neighborhood spatial correlation coefficient features,dimensionless shunting features,Gabor texture features Wavelet scale feature.And a feature filtering algorithm was applied to reduce the dimensionality of the feature library and select the subset of features with the best performance.Secondly,this paper selects five kinds of machine learning algorithms:k-nearest neighbor algorithm,naive Bayes algorithm,support vector machine algorithm,back propagation neural network algorithm,convolution neural network algorithm to classify the clutter of HFSWR,compares the classification results,selects the optimal classification method convolution neural network algorithm,and obtains a satisfactory classification effect.Finally,aiming at the problem of small samples and strong time-varying of ionospheric clutter in HFSWR clutter classification,this paper proposes a clutter classification method based on transfer learning algorithm,and further classifies the ionospheric clutter into weak ionospheric clutter and strong ionospheric clutter according to its classification results,and get a good result.