One of the most driPOrtan progresses that hav bo Inad in for techniques is theincrease in their resolutiofo which has mad it possible fOr the radar tO perfOrm not ouly theclassical taSks of detection, lOCation, searching and beking, bul Inan advanced haonsincluding target hagin and recoghon. In reCen yearS, high resolution radar targetreCOghhon has received intensive attention from the radar as well as signal ProCessingcoInInunity. Previous work put much emphasis on the develoPmen of recognition Inethods,with the data model and pre-processing muh ignOred. The woth of thes dissebon isfocused on the scatterers' model of high resolUtiOn mp pode mp) and thch feabeexthehon- The main content of this dissehation is sununarized as fonows.ChaPter l fnd gives a brief introdUCion of the research bacwi as we1l as the keytecboques of high resolution radar target recoghhon. and then provide a review ofekistingreeognihon algorithxns. Ftally the research backgrOUnd and organhation of thesdissertation is ched.ChaPter 2 establishes a staistical framwork fOr deSCribing the range Profiles andtheir nonnaled correlation, atg from the SCatterers' model. First some reasonableasSUInPtions are Inade with respect of the phase gneter of high resolution ed echoes.By utildrig the assuxnptions, the statstical propenies of scatterer-cross-tenns of the rangeprofile are established. Then. based on these proPerties, the concePtS of autoterm-tO-crossterm ratio of range profiles and of their nonnaied correlation are definrtrespeCtively, with the cormsPOnding Staisticai formula given. Finally, the staiSticalfOrmula of the nonnaiized correlation of range Profiles is established, which forms thebasis for the stability analysis of range profiles in ChaPter 3.ChaPter 3 establishes a formula of the lower bound of range profiles' nonnalizedcorrelation, starting from the statistical framework established in ChaPter 2. Theestablished lower bound explicitly expresses the range profiles' normalized correlation as ahation of aspect variation, radar signal wavelength and the target sise. Then, utilizing thelower boUnd foW4 the aspect stability of range profiles and itS dePendence on radarsignal wavelengh bandwidth and the target size are analyZed theoretically. Finally, aantable aircraft target is sdriulated, and the exPerimental resultS of range profiles' asPectstability are presented, which is in accordane wtth the conclusions drawn from thetheoretical analysis.ChaPter 4 investigates the asPect-smoothing of range profiles, Which has remained formany years the most challenging task in range profiles recognition. A new algorithm isproposed, which suppresses scatterer cross terms in the bispectral domain by utilizing their statistical properties, and then reconstruct the scatterer auto terms from accumulation ofradial slices of the smoothed bispectrum. The new method can perform averaging on range profiles without the need of range bin alignment, thus avoiding the computationalcomplexity of range cell alignment and its low accuracy of in the case of low signal-tonoise ratio or moving scatterers such as propellers. Finally, the aspect stability of smoothed range profiles are analyzed theoretically and the criterion to choose aspect interval of sample range profiles fSr averaging is given.Chapter 5 is focused on the extraction of shift invariant features of range profiles. In practical scenarios, the targets are often non-cooperative and maneuvering, as a result the positions of observed range profiles varies easily in the range window. Unless the relative position shift of range profiles are compensated, their further processing and recognition is impossible. In this chapter we propose novel shift invariant features, namely, the circularly integrated bispectra, for range profiles. The circularly integrated bispectrum preserves both the phase and scale information of rang profiles, and exhibit superior performance to the Fourier magnitudes in the respect of bo