ISAR Fast Imaging Method Based On The Multi-GPUs Parallel Optimization

Inverse synthetic aperture radar(ISAR)can achieve high-resolution imaging of the target,and can obtain relevant information such as the size and structure of the target,providing a basis for the classification and recognition of the target.With the continuous increase of the working frequency band and bandwidth of the ISAR system,the resolution of ISAR imaging results is becoming higher and higher.However,the ISAR data scale and processing time are also increasing sharply.It is difficult to meet the requirements for the ISAR imaging processing efficiency in military and civilian fields such as air defense,anti-missile,early warning and reconnaissance,etc.There is an urgent need for fine-grained parallel optimization of ISAR imaging algorithms.Graphics processor unit(GPU)has the characteristics of multi-threads and high parallelism,and its parallel computing capability far exceeds the central processing unit(CPU).Therefore,the computational complexity of the ISAR imaging process will be analyzed.The ISAR imaging motion compensation method based on multi-GPU parallel optimization will be studied.The real-time ISAR echo receiving and imaging process will be designed.The real measured data will be utilized to verify the effectiveness of the proposed parallel optimization method and the rapid imaging method.The main research contents of the thesis are presented as follows:1.discussing the basic principles of ISAR imaging and typical methods of motion compensation.Firstly,the basic principles of ISAR imaging are introduced in detail.Based on the ISAR imaging turntable model,the echo signal model using the linear frequency modulation(LFM)signal is constituted.Then,the physical meaning of motion compensation and the basic principles of commonly used envelope alignment,migration through range cells(MTRC)correction,and azimuth high-order phase compensation methods are discussed in detail.This part can lay a theoretical foundation for the following multi-GPUsbased parallel optimization.2.Aiming at the high computational complexity and low efficiency of ISAR imaging motion compensation,a fast implementation method based on multi-GPUs parallel optimization is proposed in the second part.Firstly,the GPU programming model is introduced in detail.Then,a fast implementation method of envelope alignment,MTRC correction and azimuth high-order phase compensation based on multi-GPUs parallel optimization is proposed.Finally,the calculation efficiency of the proposed GPU parallel optimization method is verified by the measured data of the Yak-42 aircraft.Compared with the CPU-based motion compensation method,the speedup ratio of traditional motion compensation methods can reach up to 191.3.To solve the problem of large data volume and low imaging efficiency when the ISAR echoes are received in real time,a multi-GPUs fast imaging method is proposed based on the thread pool architecture.Firstly,we discuss the multi-threads operation,competition conditions and methods to avoid competition.Meanwhile the thread pool construction and operating mechanism are also presented.Then,single-frame ISAR fast imaging method based on multi-GPUs parallel optimization is proposed to improve the computational efficiency of single-frame ISAR imaging.After that,the imaging processing thread pool is established when the ISAR echoes are received in real time,the ISAR fast imaging method is proposed in which different threads compete for the GPU computing resources,and reduce the resource consumption caused by the creation and destruction of GPU storage resources to maximize the utilization of GPU resources.Finally,the real-time receiving scene of ISAR echoes is simulated based on the real measured data of Yak-42.The imaging efficiency under the real-time receiving scene of ISAR echoes is verified.