Study Of Several Problems On SAR Target Recognition

Synthetic Aperture Radar(SAR) sensor plays more and more important role in military reconnaissance and achieving information dominance on the battlefield. In this dissertation, we go deep into researching on several key technologies about SAR automatic target recognition (ATR), and validate them by experiment. Firstly, the properties of SAR imagery, the both important features on grayscale and texture of SAR imagery are introduced in detail, then the unique properties and performance of the target in SAR imagery are explained, it's the basis of SAR ATR. Secondly, target detection algorithms based on const false alarm rate(CFAR) are implemented, the performances of all the different CFAR methods are compared. Thirdly, target detection and texture segmentation in SAR imagery based on fractal are deeply studied, the common methods on estimation of fractal dimension and several common features of fractal are discussed. Two improved algorithm of man-made target detection based on extend fractal (EF) with B-CFAR has been presented, the experiment's results show that texture segmentation and man-made target detection based on fractals is effective and promising. At last, according to SAR ATR system, an auto model-matching target recognition method has been presented. The experiment show it do good at recognition rate and running speed.