Inverse Synthetic Aperture Radar HRRP Target Recognition And Geometric Structure Analysis

Bacause of the characteristics of all-day,all-weather,long-range detection and highresolution imaging,inverse synthetic aperture radar(ISAR)has been widely used in military and civil fields.With the continuous development of Radar signal processing technology and the increasing complexity of observation scenes,how to use the radar observation information to achieve radar automatic target recognition(RATR)and target geometric characteristics analysis has become a research hotspot in the field of Radar signal processing.RATR technology usually realizes accurate target recognition by the preprocessing and feature extraction of Radar echoes and imaging results.Compared with two-dimensional ISAR images,high resolution range profiles(HRRP)sequence is simple and fast to acquire,which has become an important feature of RATR.Therefore,aiming at the problems of insufficient feature extraction and low recognition rate of the existing HRRP target recognition methods for wideband Radar,this thesis carries out the research of HRRP target recognition method for wideband Radar based on dual channel attention deep learning network.On this basis,the target component segmentation and geometric features,e.g.edge and contour extraction methods are also studied.The main content of this thesis is summarized as follows:1.The basic principle of ISAR imaging is discussed.Firstly,based on the equivalent turntable model,the basic principle of ISAR range-Doppler(R-D)imaging alogrithm is analyzed.The LFM-based echo signal model is also constructed.Then,the flowchart of ISAR R-D imaging is introduced,and the analytical expressions of pulse compression,intrapulse coherence compensation and translational motion compensation are derived in detail,which provide a theoretical basis for HRRP target recognition and geometric characteristics analysis.Finally,the point target model is used to simulate the algorithm in this chapter.2.The HRRP target recognition method based on Long Short-Term Memory(LSTM)-Dual Attention Module(DAM)network is proposed.Firstly,the target HRRP preprocessing method is given,and the real HRRP image sequence is obtained by energy center alignment and real envelope processing.Then,aiming at the problem that the existing deep learning methods can only extract the structure change characteristics of HRRP along the range direction,the LSTM-DAM network is proposed.Through the introduction of dual attention mechanism,the structure and attitude change characteristics of HRRP along the range direction and azimuth direction are extracted simultaneously,and the target recognition process based on LSTM-DAM is given.Finally,the effectiveness and robustness of the proposed algorithm are verified by ISAR data of aircraft targets.3.The geometric structure analysis method based on ISAR images is studied.Firstly,the target extraction methods of ISAR Image based on morphological processing,Poisson Matting and Learning Based Digital Matting(LBDM)are studied.Then,for the aircraft target and the space target,a component segmentation algorithm based on morphological processing and Radon transform is proposed to obtain the imaging results of different components of the target.Then,an ISAR image edge detection method based on Canny operator is proposed to extract the geometric structure of the target part edge.Finally,the effectiveness of the proposed method is verified by the measured ISAR images of aircraft and space targets.