Study Of Reduced-Rank STAP For Airborne Phased Array Radar

AbstractThe tecboque of Space Tfor AdaPtive Processing (STAP) can suppress cluttereftbctively and improve the detection pefformance of airborne phased array radar greatl}'.The research on STAP has been pefformed for more than 20 years at home and abroad. lnthe begindng. the fOcus is on the theory of optdrium full rank space tdrie adaptive detectionand the characteristics of airborne radar clutter. Although full rank STAP ethibits excellentperfOrmance. both its computation load and its implementation complexity are so great thatit is unable to be implemented in real time. Thereby, the focus of successive research isconcentrated on reduced-rank mp) algorithms. Once the solution is obtained. the progressto implement STAP in engineering will be accelerated greatly, which is the main puopse ofthis dissertation as well.Target detection is an issue ofbinny hyPOthesis decision. while the problem of ConstantFalse Alam Rae (CFAR) detection after STAP has been addressed in little. Two CFARdetection algorithms naxned Generalized Likelihood Raio Test (GLRT) and Adapti\.eMatched Filter(AMF) are studied in Chapter 2 and their relationships t'ith STAP arediscussed as x'ell with the results that both the AMF and GLRT statistics can be emplo}.edin CFAR processing after STAP When the size of auxiliny samples is large. the txt'odetectors nearl}' share the same performance. while the computation of WF detector is abit small.Rank reduction can be conducted in space. time or joint space time domain. Wherex'erit is reaIized. the process ofnd reduction may be equivalent to an operation on full spacetinle data 1'ectors xtith a transfer matrix Q.' If Q. is independent of the data eigen-structure- the processor structure of this kind is prefixed in adtance as well. Howe\.er. ifQ. is eigen-structure dependent. that is to sa}\ STAP is performed in data eigenspace.di11brent choices of eigenspace will result in different pefformances. The general principleof RR STAP is addressed in this paPer with the result tha the output signal to clutter pIusnoise ratio will be probably near as or even better than that of full rank STAP Thereb}. thefinal detection performance of practical airborne phased array radar won't necessarily bedegraded when rank reduction is oPerated proPerlyThere exists two basic mpes of adaPtive processor, one is Direct Form Processor (DFP )and the other is Generalthed Sidelobe Canceler (GSC). When a particular transfer matrixused to form a main channel and sevetal auxiliare channels, DFP can be converted to anequivalent GSC. The equivalence betWeen DFP and GSC is discussed in Chapter4 and theeffect of different forms of blocking matrix on STAP is analyZed in this chaPter as we1l.Study of adaPtive processing especially RR adaPtive processing based on eigenspacefacilitates to probe the optimum rank reduction scheme in theory A new algorithm namedDFP-CSM is presented in ChaPter4 based on the concept of GSC-CSM. Simulation resultsshow that DFP-CSM can aPproach the oPtimum performance more raPidly than GSC-CSMand the tWo CSM schemes can make the incurred loss of signal to clutter plus noise due torank reduction least when the rank of processor is given. Therefore, they provide thepefformance upper bounds of all RR schemes wtthin eigenspace respectively in GSC andDFP structures.In DFP structure, two eigen-based RR algorithms, i.e., Minimurn PowerEigencanceler(MPE) and Minimum Norm Eigencanceler(MNE) are discussed in ChaPter5,whose weight vectors are both confined to the noise subspace and thus the cluttercomPonent orthogonal to the noise subspace can be cancelled in nature. The intrinsicrelationship among MPE, MNE and CSM are addressed from the view of eigen-structureand their clutter suppression caPabilities are compared as well. The principal componentmethod and the minor component method are reconunended to reduce the computation ofMNE and MPE respectivelyWther rank reduction conducted o...