SAR Raw Signal Simulation And Imaging Methods Based On GPU

Synthetic Aperture Radar(SAR) is characterized by the all-weather, all-day capability to generate a high-resolution image of the targets over a long distance, plays an important role in civil and military fields. In the SAR system design and verification, imaging algorithm research, related software and hardware design, etc., we have to prepare the SAR echo signal in accordance with specific conditions. But one can't obtain it directly through the flight trial, so it is crucial to make some related study of efficient and precise SAR raw signal simulation technology. Besides, with the improvement of the complexity of the radar platform motion mode, the study of the efficient and accurate time domain imaging technology becomes important.This thesis, focusing on the background of “×××” 973 program, has made a deep study of SAR raw signal simulation and imaging methods and the parallel designs based on Graphics Processing Unit(GPU). Moreover, this thesis has systematically and deeply investigated the backward methods of SAR raw signal simulation, the forward methods of SAR raw signal simulation about 3D ground scene, Back Projection(BP) imaging algorithm, etc.. The designs based on GPU have greatly improved the operation efficiency, and have well solved the technical problem of the actual project.The main contents of this thesis are as follows:1. The characteristics of GPU and the software and hardware patterns of Compute Unified Device Architecture(CUDA) are simply introduced to make a good stage of the parallel designs on GPU. The SAR echo model and the principle of SAR imaging are discussed in detail, including the pulse compression principle of the linear frequency modulation signal, SAR spatial geometry model and the echo signal model, two-dimensional high resolution imaging principle, range-doppler algorithm, etc., which provide a theoretical basis for the full text.2. The study of the backward methods of SAR raw signal simulation is deeply made. The method of using fiducial SAR image to determine the backscattering coefficient is analyzed. Both the traditional algorithm of SAR raw signal simulation and the practical concentric algorithm in engineering are discussed. In concentric algorithm, we have made some parallel design and optimization from thread distribution, memory allocation and stream processing aspects, so as to achieve the design of double GPUs. Then, we have made some simulation experiments via a SAR image of the real scene to validate the validity of the algorithm and its parallel design. Finally, we have finished the acceleration ratio measurement of a single GPU, which is about as 193 times fast as CPU.3. The study of the forward methods of SAR raw signal simulation about the 3D ground scene is also made. The problems of using Digital Elevation Model(DEM) data to generate SAR raw signal are detailed discussed and solved.(1) The fractal interpolation method on the basis of Fractional Brownian Motion(FBM) model. We can make the DEM data fine processing to meet the requirement of SAR echo simulation.(2) The backscattering coefficient calculation method based on the facet model. We can cut the DEM data into lots of small facets whose precision meet the simulation requirement, then calculate each small facet's backscattering coefficient based on the experimental data and facet model.(3) The judgment method of shaded area based on the comparison of the overlook angles. When we treat the radar beam as an approximate sector and finished its segmentation, we can determine the shaded area with the comparison method of the overlook angles within each tiny beam slice. Their principles and specific processes in every part are detailed discussed and the corresponding simulation experiments are given. Then, we have analyzed the parallelism of this forward method of SAR echo simulation and achieved the design of double GPUs. We have also made some simulation experiments via the real DEM data of a certain region to validate the validity of this method and its parallel design. It proves that the computation efficiency of the SAR raw signal simulation about the wide range 3D ground scene has been greatly improved.4. The basic principle of BP imaging algorithm is studied. And we have also made some parallel design and optimization aimed at the problem of its low efficiency, so as to achieve the design of double GPUs. Then, some experiments via the real SAR echo data of a certain region are also carried out to verify the validity of this algorithm and its parallel design. Finally, we have also finished the acceleration ratio measurement of a single GPU, which is about 186 times faster than CPU.