CUDA-based Fast Algorithm For Electromagnetic Scattering Of Target And Environment

Up to now, electromagnetic(EM) wave has been discovered for more than one hundred years and it has been applied in many fields, such as in communications, the EM wave can carry a lot of message and propagate it at the speed of light; in remote sensing, the differences of the objeects’ scattering properties caused by the different frequency EM wave imply details of objects. For the frequency range of EM wave is large, people has been studing it for many years. And in the research of EM scattering of targets and environment, people have more focus on the scattering properties under high frequency EM wave. Because objects are usual complicated, the numerical algorithm and approximation algorithm are used rather than analytical algorithm. With the increase of frequency, it needs more memory to describe details of objects and the computational complexity gets large. Without supercomputer, it is difficult for the people computer(PC) to finish the calculation. Although many researchers have tried to combine many PC to calculate EM scattering, the result is unsatisfactory.In 2007, NVIDIA Company introduced its own parallel computing architecture, called compute unified device architecture(CUDA). It provides developers a method of using processor on Graphics Processing Unite(GPU) to calculate problems. NVIDIA designs hundreds and thounds stream processors on its GPU and these stream processors can excute threads in parallel. For stream processors are on the same GPU, its parallel performance is better than that of combination of PCs. Moreover, it provides an easy language to realize parallel program and makes it popular in many fields. The CUDA-based Fast Algorithm for EM scattering calculation of targets and environment is dicussed in the paper. The main work follows as below:First, the parallel architecture and programming model of GPU are introduced And the three models, transport theory model, four-path model and reciprocity theorem model, are programmed to exexute on GPU for calculating EM scattering proporites of three kinds of vegetation models respectively. According the differences of the three methods, three CUDA schemes are proposed to analyze problems needing attention. Compared to the general single thread CPU program, the CUDA-based programs get the speedup of two orders.Second, for the calculation of small slop approcimation(SSA) is large and hard to realize when it is applied to calculate EM scattering rough surfaces of large size, the CUDA-based SSA method is proposed to calculate the EM scattering coefficient of typical rough ground and the Doppler spectra of dynamic sea. The subsidiary thread of CPU is added and the optimization schemes of CUDA program is proposed to realize the EM scattering calculation of large rough surface. And in the research of Doppler spectra, the ship’s wake is considered in the calculation of EM scattering echo.Third, Method of Moment and the high frequency method based on geometrical optics abd physical optics(GO/PO) are combined with CUDA to calculate EM scattering of some typical targets. Especially, the effect of precision is discussed in the CUDA-based GO/PO method. As the results show, the efficient of single precision operation is higher than that of double precision operation and the single precision operation brings larger error. This problem needs to be considered in CUDA calculation.Fourth, in the simulation of synthetic aperture radar(SAR), CUDA is also used to simulate SAR echo of rough surfaces and gets a speedup of two orders. It is an efficient way to calculate the scattering echo on GPU. And the SAR images of several kinds of rough surfaces are analyzed and it is feasible to identify the rough surface from SAR images.Finally, composite scattering problem of rough surface and target is the difficult and hot spot in the EM scattering field. The hybrid method of GO/PO and SSA is proposed and combined with CUDA. The radar cross section(RCS) of rough surface with complexity target is calculated and the effect of coupled field is analyzed. With the help of CUDA, a speed up of more than one order is got. It is an effective way to calculate composite scattering problems on GPU.