Airplane Target Classification Method And Design Implementation Under Low Repetition Frequency FMCW Radar System

Since the introduction of micro-doppler effect to the radar file,the research of target recognition based on micro-motion features has attracted wide attention.The so-called micro-motion refers to the occurrence of vibration or rotation of a radar target other then the translational motion of the center of mass,or the motion of a part of a target in relation to the body.Micro-motion can produce the micro-doppler frequency.The geometrical structure and motion characteristics of different targets can produce different micro-doppler features,which is the unique character of the target.The micro-motion features play a very important role in the classification and recognition of radar automatic target.Frequency Modulation Continuous Wave(FMCW)radar obtains the target speed and distance imformation by transmitting FMCW signal,FMCW system has the advantages of simple structure,low transmitting power,and easy to realized high range resolution.When FMCW radar work on the low repetition frequency mode,the target echo will appear obvious frequency spectrum aliasing phenomenon after the frequency mixing processing.There will be a great performance loss directly using the narrow-band target classifition method suitable for Pulse-Doppler radar system.The target classification and recognition technology based on FWCW radar system has great significance.Radar signal processing is high computation cost,heavy computation and strong real-time,With the development of radar technology,the modern radar system has put forward high demands on the real-time performance of radar signal processing.In recent years,the Graphical Processor Unit(GPU)has been successfully applied in many fields.GPU has the advantages of high floating-point operation ability and high memory band-width.It is suitable for parallel acceleration problem in dealing with massive data.How to realize real-time signal processing with GPU platform has become an important research direction in the field of radar.This paper studys the problem of airplane target classification under narrow-band and low repetition frequency Liner Frequency Modulation Continuous Wave(LFMCW)radar system and radar target classification system based on GPU platform.The maincontents are as follows:1.The processing method of sawtooh LFMCW signal and symmetric triangulation LFMCW signal are introduced.The scattering point model of the airplane target rotor is derived based on sawtooh LFMCW signal.2.Aiming at the classification of three kinds of air targets(jet plane,propeller aircraft andhelicopter),the repetition frequency division and frequency mixing processing method is proposed in the narrow-band LFMCW radar system.When radar is work on the low repetition frequency mode,the traditional radar target classification method can not extract the signa of target's distance unit accurately because of the long frequency modulation cycle.So,There will be a great performance loss directly.In this paper,the echo signal will be segmented and dechirp,which is equivalent to increasing the repetition frequency of radar.The simulation experiment shows that the repetition frequency division and frequency mixing processing method can effectively improve the rotor echo of the airplane target and increase the recognition rate of the three kinds of aircraft.3.Introduce GPU and programming model of Compute Unified Device Architecture(CUDA).We analyze the difference between GPU and Central Processing Unit(CPU)in hardware,then introduce CUDA from programming model,thread and storage structure.The design of radar target classification system based on GPU is studied.Through the single instruction multiple thread(SIMT)mode of GPU thread,the parallel design of the data preprocessing method based on CLEAN and the Support Vector Machine(SVM)classification algorithm is realized.Simulation results show that the output of GPU target classification system is accurate and effective under the same processing conditions.GPU greatly improves the speed of radar signal processing and saves the time of calculation compared with the CPU platform.