Research On STAP For Airborne Radar In Krylov Subspace

Space-time adaptive processing(STAP) has emerged as a key strategic technologyfor airborne surveillance radar due to its inherent potential for significantly improvingtarget detection performance.The optimum processor,which maximizes the outputsignal-to-clutter-plus-noise(SCNR) ratio,has two major obstacles which must beovercome to make it feasible.They are the on-line computational complexity andso-called sample support problem for estimating the interference covariance matrixwhich is unknown a priori.Reduced-dimension method,which is easy to implementwith a little system performance loss,can reduce large amount of computation.Reduced-rank algorithm with complicated architecture has high performance.Reduced-rank algorithm based on Krylov subspace is signal-dependent with fastconvergence rate.The primary overall objective in this dissertation is to develop newsignal processing algorithms in Krylov subspace which advance not only the theory andperformance of STAP,but will also lead to fast implementable solutions.The main contributions of this dissertation are as follows.1.The explicit inverse of positive-definite Hermitian tridiagonal covariancematrix is relevant to the computation of algebraic cofactor thereby a fast determinantrecursion in Krylov subspace is proposed which is based on Lanczos algorithm.Thenovel method works without backward stages,therefore it is much more efficient thanMWF.Without the requirement of explicit formation and inversion of the covariancematrix,the proposed approach is a eomputationally-efficient method.Additionally,atwo-stage reduced-dimension algorithm is developed based on a data preprocessingapproach where the data preprocessor suppresses discrete interference and the adaptiveKrylov subspace algorithm suppresses clutter.The new method utilizes only a singlesnapshot of the data for adaptive processing resulting in significant computationalsavings and improved robustness to nonhomogeneous clutter scenario.2.The prior knowledge about the radar system and flight geometry can be used tospeed up the STAP algorithm with the optimum performance in Krylov subspace.Weproposed two ways to take advantage of the prior knowledge.One is to use the space-time steering vector with multiple constraints.Another is to add the space-timeconstraints to the initial weight of the recursive algorithm.3.The design of local processing region in beamspace is developed.Anonorthogonal transformation is proposed to generate the local processing region ofJDL.Selecting different nonorthogonal basis according to different Doppler frequency,the new approach improves the performance of the original JDL.The new methodfollowed by Krylov subspace algorithm can take advantage of the benefit of bothreduced-dimension and reduced-rank algorithms.4.A novel quiescent beamforming approach is proposed which imposesspace-time quadratic beampattern constraints in the adaptive processing and exploitesthe Taylor series approximation to estimate the solution resulting in low computationload.An improved STAP architecture is obtained which can obtain low sidelobeefficiently with low sample support requirement.5.Due to the time lag of adaptive weight,data mismatch may occur ifconventional STAP algorithm can't obtain notch widely enough.Based on clutterspectrum property,a new approach is presented which can widen the notch centeredaround the clutter direction therefore it is robust to the internal clutter motion.Thecorporation of the proposed method and knowledge-aided STAP can improve the totalperformance.6.Space-time generalized loading approach is proposed to improve the robustnessof STAP in Krylov subspace,which is a generalization of diagonal loading.It also cancancel clutter while maintain low sidelobe more efficiently.It has low sample supportrequirement,strong robust to error and convenient implementation.7.Space-time clutter spectrum in airborne sparse array applications is studied.Investigations show that the rank and spectrum of clutter covariance matrix aredependent on the sparse configuration and the number of temporal samples.It is pointedespecially that if the aperture do not change and the number of half-wavelengthincluded in any interelement spacing is lower than the number of temporal pulses,therank of clutter covariance matrix remain unchanged.