Browsing by Subject "independent component analysis"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    ItemOpen Access
    Methods fro automatic target classification in radar
    (2009) Eryıldırım, Abdülkadir
    Automatic target recognition (ATR) using radar is an active research area. In this thesis, we develop new automatic radar target classification methods. We focus on two specific problems: (i) Synthetic Aperture Radar (SAR) target classification and (ii)Pulse-doppler radar (PDR) target classification. SAR and PDR target classification are extensively used for ground and battlefield surveillance tasks. In the first part of the thesis, a novel descriptive feature parameter extraction method from Synthetic Aperture Radar (SAR) images is proposed. Feature extraction and classification methods which were developed to handle optical images are usually inappropriate for SAR images because of the multiplicative nature of the severe speckle noise and imaging defects. In addition, SAR images of the same object taken at different aspect angles show great differences, which makes it hard to obtain satisfactory results. Consequently, feature parameter extraction method based on two-dimensional cepstrum is proposed and its object recognition results are compared with principal component analysis (PCA) and independent component analysis (ICA) methods. The extracted feature parameters are classified using Support Vector Machines (SVMs). Experimental results are presented. In the second part of the thesis, the automatic classification experiments over ground surveillance Pulse-doppler radar echo signal are investigated in order to overcome the limitations of human operators and improve the classification accuracy. Covariance method approach is introduced for PDR echo signal classification. To the best our knowledge, the use of covariance method-based classification is not investigated in radar automatic target classification problems. Furthermore, different approaches which involves SVMs are developed. As feature parameters, cepstrum and MFCCs are used. Performances of these two approaches are compared with the Gaussian Mixture Models (GMM) based classification scheme. Experimental results and conclusions are presented.