Study On The Algorithms Of Radar Target Recognition Using High Resolution Range Profiles

The rises of modern high resolution radar provide automatic target recognition (ATR) a novel approach. High resolution range profile (HRRP) is one of the most important features with pose sensitivity. Only by further processing can it be used to effective recognition. In this dissertation, the problem of automatic recognition of radar target using high resolution range profiles is mainly studied. The contents of the dissertation mainly include the following several aspects:In chapter 1, the significance and the background of the research for radar target recognition is expounded. Then the representative algorithms for target recognition using high resolution range profiles are analysed, and two key points are indicated. Finally, the outline of the dissertation is introduced.In chapter 2, the principle of high resolution range profile and the ingredients resulted in the variability of range profiles are analysed, and the translational shift variability and pose sensibility of range profiles are emphasized in detail. Then a moment approach for extracting translation invariant features is proposed.In chapter 3, it is studied that how to select and extract mathematic features from the train samples. Firstly, the principal theory of kernel mothod is introduced briefly. Then the principles of PCA based noise-reduction and the powers of KPCA and KDDA for feature extraction are analysed in detail, and an modified KDDA is proposed. At last, a framework of automatic target recognition based on range profile and KPCA/KDDA is proposed for overcoming the aspect sensitivity of the HRRP, and the results of recognition experiment with four class airplane target is shown.In chapter 4, the model of mixtures of factor analysis is introduced firstly. Then a framework of automatic target recognition by modeling target range profile is proposed for overcoming the pose sensitivity of the HRRP. In the framework, we first estimate the probability generation model using the obtained HREP under many poses, then determine its class attribute by comparison of the conditional probability of a test HRRP. At last, the results of recognition experiment with five class airplane target using the approach above are shown.In conclusion, the main works of the dissertation are summarized and the future reseach areas are pointed out.