Research On Signal Processing Technology Of High Altitude Airborne Early Warning Radar Based On GPU

As the radar echo data that needs to be processed becomes more and more complex,and the amount of data becomes larger and larger,the real-time nature of the system becomes one of the bottlenecks in the development of radar signal processing.In recent years,with the continuous improvement of the computing performance of the Graphics Processing Unit(GPU),and the advantages of strong portability,high scalability,good reconfiguration,and decoupling of software and hardware,GPUs are used in the radar field.It has been more and more widely used.Many algorithms in radar signal processing have good parallelism and are very suitable for GPU acceleration processing.This thesis takes the GPU implementation of high-altitude airborne early warning radar signal processing as the research background.The specific work content is as follows:(1)Research on heterogeneous architecture and unified computing device architecture(CUDA)programming,analyze the hardware architecture and software architecture of GPU,and summarize and analyze some effective optimization directions in CUDA programming.(2)The signal processing flow of airborne early warning radar is studied,which mainly includes digital beam forming,pulse compression,moving target display and moving target detection,main lobe clutter suppression method,and constant false alarm detection.The basic principle of each algorithm in the processing flow is introduced,the CPU simulation realization of each algorithm is completed,and a method for accelerating the constant false alarm detection based on the integral graph algorithm is proposed.(3)The parallelism in the signal processing of airborne early warning radar is analyzed in detail,and the GPU implementation of each signal processing algorithm is designed.Combining the characteristics of moving target display and moving target detection GPU implementation,a GPU-based joint optimization is proposed.Method,and further optimize and accelerate the radar signal processing GPU program from the aspects of data transmission,thread organization structure and the use of shared memory.(4)The characteristics of the embedded GPU platform are studied,and the GPU program is further optimized and accelerated in terms of data transmission,algorithm improvement,data volume increase and instruction memory access,etc.,combined with the characteristics of the Jetson AGX Xavier platform.The test results show that the GPU The program can meet the system requirements.