Research On The Key Technology Of Radar Azimuth Super-resolution

This paper analyses different azimuth super-resolution methods and summarizes the superiority and limitations in the existing methods.The solutions and improved methods are presented for the proplems of those super-resolution methods.Meanwhile,the simulation results of these improved methods are provided under the specific experimental condition.The de-convolution azimuth super-resolution theory is elaborated the by modeling radar echo signal.Meanwhile,it analyses the results of these existing de-convolution methods,including wiener filtering method,Richardson-Lucy(RL)method and maximum a posteriori(MAP)method,the simulations of which are presentd only considering the niose in the receiver.Because the systematic noise is unknown and has a great influence on the resolution performance,there is an "ill-posed" characteristic in de-convolution super-resolution problem.In the first place,gradient projection(GP)azimuth super-resolution method based on norm regularization is proposed to address that ill-posed problem by converting convolution azimuth scaning model into regulalrizaton model.Consequently,target azimuth signals are distinguished while signal to noise ratio(SNR)equals 0 dB.Dual-logarithmic barrier super-reolution method is presented by means of alysing GP method.It depends on the dual theory and the logarithmic barrier function and the resolution ration can be enhanced successfully under a 20dB-SNR circumstance.In the second place,two methods on account of the quadratic programming are proposed for increasing resolution ratio of regularization methods,namely fast iterative shrinkage thresholding(FIST)and Lagrange method.FIST is used for improving the resolution ratio of GP methods,and Lagrange method is used for solving the problem that regulalrization parameter can not be determined adaptively.Consequently,the initial azimuth signals are restored by Lagrange method without regulalrization parameter.Two de-correlation estimation methods are proposed for improving accuracy ratio of coherent signals' DOAs estimation,including SVD-FOC method and OR-MUSIC method.SVD-FOC method is combined singular value decomposition(SVD)with fourth-order cumulant(FOC)that coherent sources are estimated successfully at low SNR.OR-MUSIC is proposed relying on the compressive sensing theory.It is applicable to large scale antenna array that when reconstructed expression of steering vector is deduced with orthogonal matching pursuit(OMP)algorithm.Consequently,the DOAs of coherent sources are resoundingly estimated by subspace algorithm and then the validity of this method is illustrated with several simulations.