Research On Ground Target Identification Through Radar System

With the development of artificial intelligence and other relevant technologies,Advanced Driver Assistant System(ADAS)and Automatic Driving System(DAS)have become one of the research hotspots all over the world in recent years.Radar is very important in ADAS and DAS due to its stable performance in adverse weathers,and the research of ground target identification through radar is of high scientific significance and engineering value.By analyzing the echo from target,radar based sensing technique is able to realize the target detection,localization and identification.Several research areas such as beamforming,target identification,and attenuation/scattering analysis are involved in radar sensing.Although these techniques have been investigated for many years,they are still open research questions and need further investigations.On the other hand,despite the development and prevalence of millimeter wave radar,the attenuation and scattering behavior of millimeter wave have not been conducted a thoroughly research,which impedes the application of these systems in outdoor environments.To solve these problems,intensive research of fixed wideband beamforing,ground target micro Doppler signature modeling,target identification through micro Doppler signature and millimeter wave attenuation has been done in this thesis.The main innovative achievements in this dissertation include:(1)A fixed wideband beamformer with frequency invariant property is proposed.By taking the Response Variation(RV)into consideration during object function construction,a fixed wideband beamformer with frequency invariant property can be achieved.Least suqare method and eigenfilter method are adopted as the optimization method to take advantages of their low computation burden.The eigenfilter method is solved by transforming the object function into a generalized eigenvalue problem.Furthermore,we also investigate the null designing through linear constraint method and null's energy expansion method when the beamformer is located in a environment with interfere appears.Simulation results show the feasibility of proposed fixed wideband beamformer with changing parameters.(2)Analysis of pedestrian's micro Doppler signature modeling is conducted and a model for micro Doppler signature of bicyclist is proposed.The Radar Cross Sections(RCS)of pedestrian with different aspect angles and RCSs of pedestrian's body segments at 300 GHz are measured.Then,the simulation and experimental measurement of micro Doppler signature of pedestrian are done at 24 and 292 GHz to show the advantage of the result at 292 GHz.Moreover,a model can be used to simulate the micro Doppler signature of bicyclist is proposed and validated with indoor experiments at 292 GHz and outdoor experiments at 24 GHz,respectively.The experimental results indicate that better continuity can be found at the micro Doppler signature at 292 GHz and the proposed model can be adopted to simulate bicyclist' micro Doppler signature properly.(3)One feature extractor for micro Doppler signature based on sparse coding and two improved signatures are proposed.A sparse coding based feature extraction method for micro Doppler signature is presented to investigate the identification of ground target with different aspect angles,signal dwell time,etc.Based on the combination of range profile and micro Doppler signature,range micro Doppler signature is proposed as a potential solution for the decline of identification accuracy when target's moving direction is normal to radar's Line of Sight(LOS).Range micro Doppler signature is able to capture the weak components in traditional signature,enrich the classifier's input information and enhance the accuracy.Furthermore,tangential micro Doppler signature is proposed to record target's tangential moving information which reach its maximum when target's moving direction is normal to radar's LOS.The experiments show that the proposed feature extractor performs better than other feature extractors with different aspect angles as well as signal lengths.The classification results show that the identification though range micro Doppler signature can reach higher accuracy when target's moving direction is normal to radar's LOS.The simulation results indicate that tangential micro Doppler signature can be a potential solution to improve the identification accuracy.Furthermore,intensitve experiments have been conducted to investigate the attenuation produced by radome contaminations on millimeter waves in outdoor uncontrollable environment.The attenuation model based on Fresnel reflection/refraction function and transmission line theory is validated by menasuing the attenuation produced by increasing contaminations.The radome contaminated media includes pure water,rain water,salty water,ice,sand,etc.A three-layer media model is adopted,and the corresponding attenuation model is proposed based on the combination of Fresnel reflection/refraction function.The experimental results show that the model can be used to analyze the attenuation precisely and the attenuation produced by radome contaminations is limited.