The Design And Implementation Of Radar System Based On GPU And CPU

With the high-speed development of the technology and owned economy, thedata processing ability of the radar system that on the strength of PC as the hostmachine has been greatly improved. It is a new trend that the current advanced radarterminals become generalized, informationized, softwaralized, which has taken greatconcern and attention of related fields at home and abroad.Traditional radar terminal systems that adopt CPU architecture, has faced manybig challenges from cost, performance, accuracy and real-time performance. TheDifferent structure system based on GPU and CPU, which incorporates theadvantages of the general processor and the high speed processor, has become amainstream in this field as to cost boost and performance improvement.With the huge datas and high real-time performance requirements of the Radar signalprocessing system, the hardware difficulty has become a key factor that limits itsdevelopment. The Graphics Processing Unit has many computing units, with apowerful parallel computing capabilities and high computing efficiency, and hasbrought new opportunities to the classic radar signal processing. Based on the idea, asoftware radar terminal system based on a GPU and CPU is developed in order toachieve a practical and high-performance informationalized radar terminal. The mainresearch achievements are as follows:(1)The development of radar terminal based on GPU+CPU heterogeneoussystem was conducted, improving radar data and signal process from softwareterminal，implementing the generalized and softwaralized radar terminal, whichreduces the usage rate of a CPU to less than11%. A software product has already gota real radar data test and been validated from demanders.(2) A radar terminal’s the signal processing of CFAR algorithm for parallelimplementation has been studied and the simulation of terminal experiment isfinished. When data volume reached a certain number, the operation efficiency is fivetimes more than the CPU’s. Thus it can significantly increase the real-timeperformance of radar signal processing.(3)The polar coordinates transformation based on GPU is implemented and the issues of information lost in the conversion process for the polar coordinates aresolved, the display of more accurate and complete. There are different numbers ofpulses in per frame, while you must follow the minimum number of pulses per framewhen generating a mapping file in a traditional way, which results in the loss of someimportant information.