Research On Time-Frequency Feature Extraction Of Dual-Channel SAR Micro-Motion Target

When synthetic aperture radar(SAR)works,it will encounter a special class of targets,such as rotorcraft or vibrating mechanical equipment.There are many very complex nonlinear components in the echoes of these targets,which are called micro motion targets.Through the study of these nonlinear components,the key information on the battlefield,such as the physical structure size and motion information of the detected target,can be analyzed.It has rich applications in battlefield situation awareness,target detection and recognition,and is the frontier issue of SAR research at present.Based on the signal modeling of dual channel SAR,this dissertation deeply studies the parameter estimation of single micro moving target,the feature extraction of micro moving target signal under the condition of time-frequency imperfection,and the application of neural network method in micro moving target detection.In Chapter 2,the motion of micro moving target and SAR echo signal are modeled.After decomposition,the motion of the micro moving target can be expressed as sinusoidal motion on the two motion axes,so the echo signal also shows the characteristics of sinusoidal modulation.In order to improve the anti-interference ability of the signal,the echo can be re expressed as a sinusoidal amplitude modulated signal by using the phase center offset antenna(DPCA)technology,and the parameters of the virtual image are given.In Chapter 3,the parameter estimation and target detection algorithm of micro moving target are studied.Based on the principle of stationary phase and by locating the maximum value of the mixed modulation signal,an effective method for estimating the sinusoidal frequency modulation signal under the condition of low signal-to-noise ratio is proposed.At the same time,the estimation accuracy can be improved by using time smoothing and quasi maximum likelihood methods.The simulation results show that the method is effective under the condition of low signal-to-noise ratio.In Chapter 4,when the time-frequency is incomplete,the echo signal of the micro moving target will be seriously missing in time domain or frequency domain.The time-frequency characteristic estimation formula of complex micro moving target is given by using Viterbi algorithm.The simulation results show that it has good performance in low signal noise.On this basis,the time-frequency features of multiple incomplete micro moving targets are extracted and estimated by using peak location and bidirectional addressing methods.The experimental results show the effectiveness of the algorithm.In Chapter 5,the application of neural network in micro motion target detection is explored,and the automatic structure search(NAS)is mainly studied.A new NAS mechanism,called balanced NAS framework,is designed as an alternative to the traditional method of fixing neural structure.Its core is to propose a balance function to manage NAS,so as to meet the real-time requirements of micro motion target detection,and promote the neural network in micro motion target detection to have high accuracy and efficiency at the same time.The experimental results show that the "block cell" cell structure and the new NAS mechanism of RNN controller can be used to design a model equivalent to the manual network design.In this dissertation,the time-frequency feature extraction of micro moving targets in many complex situations is considered,which has certain guiding significance for SAR micro moving target detection closer to the actual battlefield.