Research On Reconstruction And Target Recognition Of Missing Signal Of Radar High Resolution Range Profile

High Resolution Range Profile(HRRP)contains abundant target structure information,which is of great significance for radar target recognition.However,the actual radar often works in a complex electromagnetic environment.At this time,due to the influence of various interference and other factors,some frequency points of the obtained HRRP signal may be missing,The direct use of this missing signal will lead to the serious decline of radar target recognition performance.In recent years,deep learning model has attracted extensive attention because of its superior feature extraction ability,and has gradually been widely used in the field of radar target recognition.This thesis will study the reconstruction and target recognition method of HRRP missing signal based on deep learning model.The main work of this thesis is summarized as follows:(1)The second chapter first introduces the basic concept of HRRP and its sensitive problems and solutions;Then the traditional HRRP target recognition method is introduced;Finally,the traditional missing signal reconstruction method is introduced.(2)In chapter 3,aiming at the problem that the missing frequency points in HRRP lead to the decline of radar recognition performance,the HRRP missing signal reconstruction methods based on Bidirectional Recurrent Imputation for Time Series(BRITS)and Multidirectional Recurrent Neural Network(MRNN)are studied respectively,Based on MRNN,a Multi-head Self-attention Multi-directional Recurrent Neural Network(MSMRNN)is proposed to reconstruct the missing signal of HRRP.The model introduces the multi-head self-attention mechanism to solve the long-term dependence problem of MRNN model.At the same time,the multi head self attention mechanism is used to select the features extracted from MRNN to reduce the interference of redundant information.In addition,MSMRNN model also introduces the temporal decay factor in BRITS model to control the information flow between hidden layers,so as to reduce the impact of the historical information of a feature on the model and improve the performance of the model when a feature is missing for a long time.Experimental results show that the proposed method can obtain better reconstruction performance,and the reconstructed HRRP signal can significantly improve the recognition rate.(3)In this paper,Transformer-AM model is proposed to realize end-to-end target recognition of HRRP missing signal.The model adds an attention layer and softmax layer to the encoder group of transformer,and then classifies the missing signal by inputting the encoder group output corresponding to the observed value in the HRRP missing signal into the attention layer and softmax layer in turn.In addition,Transformer-AM performs mask operation in the multi-head attention layer to avoid the impact of missing values on the recognition results.The experimental results show that compared with the method of reconstruction before recognition,the model can further improve the target recognition performance of HRRP missing signals.