Parallel Optimization And Implementation Of Synthetic Aperture Radar Imaging Algorithm

Synthetic Aperture Radar(SAR) have all-weather capability and high-resolution earth observation, and is widely used in battlefield surveillance and disaster relief and other military and civilian fields. SAR requires different platform as satellites and aircraft and different imaging conditions as positive side and strabismus and centimeter-level resolution, it makes SAR imaging algorithm showed exhibits complex and data quantification trends. This changes presented a huge challenge to quasi-real-time SAR imaging processing platform, so parallel optimization SAR imaging processing algorithms based on high-performance computing platform become to be a research hotspot.This article studied a parallel optimization research about positive side and strabismus SAR imaging processing based on high-performance computing architecture using Graphics Processor Unit(GPU), and build a SAR imaging processing simulation system based on GPU which had verified the functionality and performance optimization algorithms. Research and experimental results show that GPU-based parallel computing architecture and optimization methods can significantly improve the performance of a typical application background SAR imaging processing algorithms and achieved a quasireal-time image processing.This article studied Compute Unified Device Architecture(CUDA) parallel computing model based on analysis about SAR imaging processing technology and its development trend and this laid the technical foundation for the further optimization of imaging algorithms.Studied the positive side SAR imaging principle and Range Doppler(RD) imaging algorithm, and divide the RD algorithm into two parts according to GPU and CPU different computing features to solve the problems that interpolation operation cannot be multi-threaded parallel computing. Range compression and distance data interpolation parameter calculation is calculated by CPU. Interpolation according to the interpolation parameters and azimuth data compression is calculated be GPU. This method divided the RD rational to GPU and CPU full use the GPU and CPU computing power and their characteristics, effectively improve the computational efficiency of imaging algorithms, and meet the quasi-real-time computing needs.Studied the strabismus SAR imaging principle and the Chirp Scaling(CS) imaging algorithm. Conducted a detailed theoretical analysis about the different range migration correction between small and large squint. Then reasonable allocation of CPU memory and GPU memory to complete the assignment combining the characteristics of CS algorithm, this make the CS parallel algorithm based on GPU has a quasi-real-time computational efficiency.Also studied the SAR echo simulation method and performed echo simulation algorithm parallel optimization. Proposed a SAR efficient simulation method based on echo delay discretization. Build a SAR simulation system which takes full advantage of GPU and CPU. This system verify the performance of the imaging algorithm provides high-performance simulation environment.Conducted several experimental under different scenarios and various imaging modalities and verify the functionality and performance of parallel imaging algorithms. Simulation results show that GPU-based parallel imaging platform SAR quasi-real-time performance, and have a wide range of engineering applications.