Design And Implementation Of FMCW Millimeter Wave Radar Road Target Classification System

With the rapid development of urban transportation and the increase in the number of vehicles,the construction of smart cities and smart transportation has also been strengthened.Millimeter wave traffic radar has become an emerging traffic monitoring technology due to its advantages of high accuracy,all-weather,high safety,and low energy consumption.Target recognition and classification technology utilizes the characteristics of road targets to achieve recognition of target types,which can improve traffic safety and optimize traffic management.It is an important technology in the field of millimeter wave traffic radar.However,traditional target classification methods face many challenges in complex traffic scenarios,including low number of point clouds,occlusion,and poor recognition performance of weak targets.Therefore,conducting classification research on millimeter wave radar road targets can improve the performance of traffic radar and promote its scientific research and application development.This article discusses the relevant principles of FMCW millimeter wave radar target detection and estimation,analyzes the content and extraction methods of road target features,and proposes an optimized classification method based on convolutional neural networks to solve the problem of poor classification performance of traditional traffic radar targets.This article has done the following two aspects of work:(1)Study the FMCW millimeter wave radar road target classification method,analyze the feasibility of feature input classification method and CNN based automatic classification method,and explore the content and methods of feature extraction.Exploring classification methods for highresolution and medium resolution systems using CNN,and studying the characteristics of HRRP and MRRP.In response to the low accuracy of MRRP in identifying weak targets,an optimization method is proposed,which is to use Joint Range Profiles of Continuous Doppler(JRPCD)to improve the accuracy.(2)Based on the above research,a road target classification system based on FMCW millimeter wave radar has been implemented.Provide a general design and module design scheme for the system,detailing the design ideas and implementation functions of the millimeter wave radar sensor,data acquisition,data preprocessing,and target classification modules.Finally,using multiple datasets collected by the system and combining research on two classification methods for road targets,the recognition and classification of FMCW mmwave radar road targets were achieved.Among them,the feature input method analyzed the strong dependence between prediction performance and velocity characteristics;The MRRP optimization method was validated using a classification method based on convolutional neural networks.Two structural modes of JRPCD(J-MRRP and M2D-MRRP)were experimentally compared to improve the recognition accuracy of low intensity targets.