Implementation Of High-speed Maneuvering Target Detection Based On Blade Server

Traditional radar systems are developed using proprietary hardware architectures such as DSP and FPGA.This development mode has a long development cycle,low scalability and poor reconfigurability.Therefore,multi-purpose software radar development platform can be realized by reconfiguring RF hardware and programmable software.Due to the high flexibility of the software platform,developers can perform verification tests on various radar signal processing modules through software programming.Traditional radar can only be used for specific tasks,and the hardware of software radar has the characteristics of reusability,and it can speed up the development cycle and reduce the production cost.In this paper,the echo model of high-speed maneuvering target is established first,and its echo characteristics are analyzed.At the same time,the effects of maneuvering target speed and acceleration on matched filtering are studied respectively.For the distance walking problem of high-speed maneuvering targets,Key Stone transform compensation,frequency domain phase compensation,and RFT transform compensation are used to correct the target of echo signal on the same distance unit,and the advantages and disadvantages of the three in engineering implementation and computational complexity are compared.Aiming at the Doppler expansion problem of high-speed maneuvering targets,the acceleration compensation of the target signal is carried out by the Dechirp method,the FRFT method and the delay compensation method,so that the energy obtained by the coherent accumulation is maximized.Secondly,this paper builds a strong real-time high-speed maneuvering target signal processing platform on a general-purpose platform blade server,and designs a multi-channel data parallel,pipeline parallel,and thread pool parallel radar signal processing framework.The processing framework can be formulated according to actual needs.For the front-end of system data acquisition,the system develops FPGA-based PCIe signal data acquisition board and host computer driver module,and sends radar echo data into server memory for processing through optical fiber and PCIe bus.For the back end of the system,this paper decouples the algorithm of high-speed maneuvering target signal processing into multiple independent modules,and performs verification and real-time test optimization respectively.At the same time,this paper designs a pulse compression component based on OpenMP;proposes a method of using multi-thread parallel segmentation to search for target speed to reduce the system calculation delay;optimizes the calculation process for each compensation algorithm module and reduces memory allocation resources;plots centroid algorithm method based on pre-sort and hash table is proposed.By planning a reasonable data flow,the entire system divides the multi-threaded tasks appropriately,constructs each signal processing module to meet the actual project requirements.The system can change and upgrade the signal processing scheme according to various radar application requirements,which greatly improves the system of scalability.Finally,this paper optimizes the real-time performance of the software radar system built on the blade server platform,and conducts a standardized design at the software code level to facilitate the maintenance and upgrade of subsequent developers;detects the performance bottleneck of the program to optimize the program performance;analyzed the performance of multi-core CPUs and optimized thread scheduling;designed a memory pool to reduce memory fragmentation and speed up memory allocation,thereby reducing the computing delay of the system;embedded real-time patches into the system operating environment to enhance the real-time performance of the system.In this paper,the entire software-based signal processing process is built on the multi-core CPU platform.From data inflow to generation of dot trace information,reasonable design and optimization are carried out,and multiple CPUs are fully utilized to meet the real-time requirements of the project.