Research On ISAR Rapid Imaging System

With the realization and development of two-dimensional high-resolution inverse synthetic aperture radar imaging technology,the research on high-resolution ISAR rapid imaging technology becomes crucial.For ISAR imaging at large rotation angles,the target translation and rotation are often coupled together.If the traditional ISAR imaging algorithm is still used,it will cause image defocus distortion and affect the imaging result.Therefore,the polar coordinate format algorithm can be used to improve the ability of phase compensation,and combined with the minimum entropy autofocus method to extract the error phase,the algorithm not only has the imaging ability to cope with the large rotation of the target,but also has the characteristics of high efficiency,fast calculation and small amount of calculation.In addition,the ISAR imaging technology often reflects the high concurrency of the signal level.If only relying on the traditional serial computing method for single-core or multi-core CPU imaging,it will be difficult to deal with the huge calculation amount of the ISAR echo signal in a short time.The GPU is composed of thousands of small and efficient hardware cores,and its high-speed general-purpose computing performance provides the possibility for the acceleration of the ISAR rapid imaging system.Therefore,the radar imaging parallel acceleration technology on the CPU+GPU heterogeneous platform Came into being.The main works of the thesis are as follows:1.Aiming at the problem of coupling of translation and rotation of ISAR imaging under large rotation angle of the target,the polar coordinate format algorithm with less calculation is used for error compensation,and the minimum entropy autofocus method is used to extract the error phase and compensate,then improving the imaging effect.The feasibility of the scheme was verified by the complete ISAR imaging algorithm simulation experiment,and the parallelism of the algorithm was obtained through derivation of the formula.2.Aiming at the problem of huge calculation amount of phase compensation during ISAR imaging processing,a GPU accelerated parallelization scheme for phase compensation is proposed.In the CUDA programming model,the memory model design and kernel function design are carried out,and the CPU+GPU heterogeneous parallel acceleration scheme is proposed for the estimation of phase error and the minimum entropy phase compensation method.By analyzing the performance test results of Nsight/Profiler software,the CUDA kernel function is optimized for the instruction delay problem in program execution,effectively hiding the delay effect,and doubling the program efficiency.Aiming at the problem of tedious calculation and low efficiency caused by the frequent use of FFTshift and FFT in the minimum entropy autofocus phase compensation method,a combined time/spectrum shift conversion method is proposed,which reduces the complexity of the code,improves the function integration,and improves Parallel efficiency of ISAR phase compensation function.It also provides ideas and methods for the parallelization of the ISAR rapid imaging system.3.Combined with the research on the signal level concurrency of each module in the ISAR imaging process,the PARALLEL design of ISAR imaging algorithm under target large Angle motion is carried out by using CPU+GPU heterogeneous parallel development technology.The parallel programming experiment is completed by CUDA technology simulation,and the feasibility of the scheme is verified through experiments.In this scheme,GPU resources are fully utilized,threads are configured,more than 60 kernel functions are called,memory is rationally allocated,access time and running time are shortened,and the parallel computing effect of CPU + GPU heterogeneous platform is more than 60 times faster than that of serial executed ISAR imaging algorithm.At the same time,analysis and experiments were conducted on the difference in computing power of different floating-point types.It was verified that the single-precision computing power of the parallel experiment in the GPU is far superior to the double-precision computing power,and the parallelization of the ISAR imaging system was optimized to further enhance the acceleration effect.The acceleration effect is more than 110 times faster than the CPU platform.