ISAR Fast Imaging Method Based On The Platform Of GPU

Inverse synthetic aperture radar(ISAR)imaging plays an important role in space situation surveillance and air defense anti-missile areas since it can not only obtain high-resolution target image but also offer unique and distinct characteristics of target,such as size,structure and etc.It provides evidences for target classification and identification.After the development over than 60 years,the basic theory of ISAR imaging have been well formulated.However,with the update of radar hardware system and the increasing requirement of target observation,ISAR signal frequency transforms towards highfrequency band and radar signal bandwidth becomes broader,which results in a rapid expansion of the radar echo data scale,so that the computing speed of ISAR imaging slows down.To accelerate ISAR imaging computing speed,this thesis makes a research on Graphics Processor Unit(GPU)based ISAR fast imaging method,which achieves the ISAR real-time imaging by taking advantage of GPU's multicore structure.The content of the thesis is summarized as the following three parts:The first part introduces the basic theory of ISAR imaging and establishes the geometry model and signal model of ISAR imaging.To lay a foundation for GPU based ISAR imaging,the physical meaning of translational motion compensation is illustrated and clarified,also the common used method of range alignment and fast minimum entropy phase compensation method are discussed.The second part proposes a GPU based minimum entropy phase compensation method to solve the inefficiency of ISAR imaging processing.The compute unified device architecture(CUDA)programming model is introduced firstly.Then,a CUDA based fast Fourier transform(FFT)method,which is widely used in minimum entropy phase compensation,is proposed.Through reasonable data distribution and memory utilization,this method obtains 80 x speed-up ratio by comparison to use central processing unit(CPU)only.Next,a CUDA based parallel accumulation method is introduced,which achieves 60x-90 x speed-up ratio by comparison to use CPU only in one-dimensional and two-dimensional accumulation computing.Finally,combined with parallel FFT and parallel accumulation method,a GPU based minimum entropy phase compensation method is proposed.This method achieves at most 70 x speed-up ratio by a new computation flow which is designed for CUDA execution model.The third part proposes a parallel ISAR fast imaging method to maximum the efficiency of CPU and GPU.At first,a multithreading computing model is introduced.Next,the implementation of matched filtering and range alignment in ISAR is presented.In order to solve the problem of process control and task allocation,this part suggests a solution which uses multithreading to process different echo data in parallel,and each thread use state machine to control the process flow of ISAR imaging.A mutex is used to ensure only one thread can access the GPU at any moment.Finally,this solution gains 50 x speed-up ratio by comparison to use CPU only.When processing the echo data with 256 points at azimuth and 512 points at range,44 frames can be obtained in one second.