Fast Simulation Of Target Feature For Synthetic Aperture Radar

As a microwave remote sensing tool, Sythetic Aperture Rader has a all-sunlight and all-weather feature. It has great potential applications in the fields of military and civilian. Because of the microwave imagery is not intuitionistic. It is hard to identify targets from SAR images by people or machine. The simulation of specifical target's raw signal and image in given condition, which expand a new way to collect target samples,to predict target features, and to implement target recognition.Based on high frequency predict method, we adopt two kinds of simulation model to simulate target feature. The first model is accurate and the second one is approximate. For the accurate model, we firstly compute the wide-band RCS curve, then use multiple frequency-point imaging method to get the one dimension range profile,lastly adopt accurate imaging model to generate SAR simulation image. For the approximate model,we firstly compute the distributing of electromagnetic field in the central frequency and central position,secondly convolute the scattering strengeth with the system transmission function and lastly map the convolution result into the imaging array to get the approximate simulation image of SAR. We bring forward the concept of normalized unit scattering strength (NUSS) to consummate the theoretical model of approximate method. Normalized scattering strength represent the contribution of single surface element and edge element to the scattered field, and reflect the distribution of scattered field.Based on the statistics theory of radar scattering, we adopt facet model to simulate distributing scene of ground. In the case of using square-law detection, surface scattering strengeth is obtained by a random number of exponential distribution, and its mean value is determined by physical optical theory. Finally we use mapping method to obtain the background of SAR image.We adopt the real-time program ability of Graphics Processing Unit (GPU) to improve the Graphical Electromagnetic Computing (GRECO) method. Based on the spatiality and parallelism of the computation of SAR simulation, we transfer the computation work of scattering field analyze and SAR image formation from CPU to GPU. The program is coded by Cg Language (Graphics C Language). The mainly functions of the code is to changes the work flow of fragment processor to complete the computation of electromagnetic analyze and image formation.All of the algorithms involved in this thesis are integrated into a independent simulation system. We demos some results generated by the simulation system on the conditions of various imaging modes and various situations. The simulation data show high visual similarity.