Research Of Airborne Synthetic Aperture Radar Echo Data Simulation Based On GPU

Nowadays, synthetic aperture radar(SAR) technology is widely used in military and other fields, SAR echo simulation technology becomes more and more necessary and important. In previous studies, large amounts of data and calculation were great challenges and bottlenecks of echo simulation research. With the emergence of more diverse simulation platform, echo simulation has a new breakthrough point in improving the computation speed and efficiency. In recent years, GPU and CUDA programming platforms show up, making GPU-based parallel programming begin to be used in various fields. Researchers mixed CPU and GPU technology to program heterogeneously, making the program to achieve higher computing performence. Study found a large number of suitable parallel computing part in SAR echo simulation, one can use the powerful architecture of the GPU and floating point capability of parallel computing to accelerate. This thesis focuses on the GPU acceleration and optimization of SAR echo simulation.The main contents are as follows:1. This thesis introduces the basic SAR echo simulation theory, models and echo imaging algorithm, and also analyzes the models and methods used in this study. Then proposes the basic optimization strategy based on common GPU model and CUDA programming model.2. This thesis uses the echo time-domain simulation algorithm based on forward method. The validity in MATLAB is verified firstly, and then point target and lattice target of echo imaging on the CPU and GPU are realized and verified. The speed of point target echo simulation on the GPU is 410 times that of CPU.3. An acceleration strategy based on forward method on GPU is Proposed:at first, optimize processes, reduce the data transfer between the CPU and GPU; To accelerate the parallel block calculation on GPU, block size adaptive classification method is proposed.Base on the method,suitable simulation environment and the algorithm of optimal block size classification for this paper is found; For the multiple loops in operation, this thesis analyzes different efficiency of coarse-grained differentiate, choose coarse-grained division; Aimed at the problem of low efficiency of 2D echo data access, linearizes the 2D data. After the optimization strategy to accelerate the echo simulation on the GPU, acceleration ratio increased to about 480 times.4. This thesis using inverse CS echo simulation method based on inverse method, echo imaging simulation is verified on both CPU and GPU. Focusing on the specific implementation process of echo simulation, this paper extracts parallel section and calculate in the GPU block by block. Then uses the library on the GPU to compute FFT, analyze and select the optimal matrix transpose method, optimizes to reduce low efficiency parts such as direction. Due to these optimization strategies, the speed of computing optimized reaches 4.7 times that of CPU.