The Fixed-point Fast Simulation Technology Of SAR Real-time Imaging At System-level

Synthetic aperture radar is a major breakthrough in the history of radar technology,with the advantages of all-weather, all-time, long-distance and high-resolution imaging etc.So SAR is widely used in military and civil fields. In recent years, with the rapiddevelopment of digital devices, the SAR real-time imaging processing technology based onhigh-speed digital devices is an important development direction in the field of radar signalprocessing. However, the implementation of embedded real-time SAR processing system isfacing with new challenges of high resolution, platform miniaturization, light-duty andlow-power consumption. Therefore, under the condition of high resolution, how to reducethe storage space and computing resources, improve the computing speed is the currenturgent task, and the fixed-point processing is an effective way to solve this issue.This paper studies both from the fixed-point processing and simulation verification ofimaging algorithm. The first chapter describes the overview of SAR, real-time imageprocessing technology and fixed-point technique. Then the second chapter introduces thetheory of the SAR imaging, RD and CS algorithm. Through analyzing the amount ofcomputation, this paper points out that the FFT is the main computation of imagingalgorithms, so fixed-point for it can accelerate the whole process, is of great theoretical andpractical significance. The third chapter finally chooses the SystemC to realize thesystem-level accurate and fast fixed-point simulation by compareing the typicalsystem-level design language SystemVerilog and SystemC difference in the aspect ofsystem-level modeling, and commonly used fixed-point simulation methods of C/C++,Matlab/Simulink and SystemC advantages and disadvantages in terms of speed, precision,operational. The fourth chapter focuses on the based-22FFT algorithm introduction.Through analysis, the paper determines the reasonable scale strategy, refines the modelwhich is divided into fixed-point and floating-point operations two parts to realizerespectively, and then presents the simulation results. The fifth chapter takes CS algorithmas an example to achieve the modeling combined floating-point and fixed-point modeling.By fast simulation of different data, according to the image quality evaluation index, thepaper analyzes imaging results from multiple aspects, and then draws the conclusion: Whenthe system word length is24bits, the simulation results meet the requirements of systemindicators.