Array Radar Signal Processing And Implementation Based On GPU

Digital Array Radar(DAR)can flexibly generate multi-beam at the same time.It has been widely used due to its advantages such as suppression of interference,low probability of interception,simultaneous multi-function implementation,and high scalability,etc.With the development of radar technology,the data size of radar echo and the computational complexity of the algorithm have increased year by year.So high-speed signal processors and high-efficiency parallel algorithms are the very important issues in the radar field.The conventional radar signal processor is mainly based on FPGA and DSP embedded system,while GPU has prominent advantages in general purpose parallel computing.Thus,when developing a certain digital array radar experimental system processor,focusing on the difficulty of Multi-beam,multi-task real-time monitoring,this work analyzed and summarized the implementation of parallel DAR algorithms based on GPU.First,this thesis introduced the signal processing principle of phased array radar.In consideration of engineering requirements,the basic flow of phased array radar signal processing is given.According to the processing flow,this work made a further research about the basic principles of digital beam forming,pulse compression and moving target detection,etc.Further more,A new design method of MTD filter based on second-order cone optimization is proposed.Meanwhile,this thesis focused on the basic concept of MIMO radar and signal processing flow of MIMO radar,and gave a detailed account of differences between MIMO and phased array radar during signal processing flow.Then,this thesis focused the hardware architecture of GPU,CUDA kernel,the structure of thread and storage,software system and programming model of CUDA,which provided the theory foundation for the subsequent research of signal processing platform.The signal processing system adopted the modular heterogeneous parallel architecture and used NVIDIA K40 GPU as the basic computing platform.Finally,this thesis studied the radar signal processing and its implementation based on GPU.On the basis of radar's function and performance requirements,the framework of signal processing software system and algorithm module is constructed.And a specific implementation method of digital array radar signal processing algorithm on CPU+GPU collaborative heterogeneous parallel platform was designed.In this thesis,four key algorithms: pulse compression method,constant false alarm detection method,point trace condensation and target angle measurement,are designed and optimized parallelly on GPU platform.The stability,correctness and acceleration of the parallel algorithm are verified in the performance comparison experiment with the traditional embedded platform.Finally,the design and test of the radar processor system is successfully completed,and the result of experimental shows that the system has good real-time performance,parallelism and expansibility.