Research Of On-Vehicle Millimeter Wave Radar Target Detection Technology

With the innovation of vehicle detection technology and the rapid development of the automotive industry,automobiles have become the main means of transportation for human travel and cargo transportation.The urgent needs of vehicle control and safe driving put forward higher requirements on the performance and environmental adaptability of on-board sensors.Millimeter-wave radar has become the core vehicle-mounted sensor that supports assistant driving or even autonomous driving due to its high accuracy,high integration,low cost,and its ability to work around the clock and all day.The target detection technology of millimeter wave radar in complex environment has gradually become a research hotspot.The thesis focuses on the technical problems of multi-target reliable target detection for vehicle millimeter-wave radar in complex environments,and carries out research on vehicle millimeter-wave radar target detection and distance,speed,angle and other parameter measurement technologies.In this paper,the research of target detection technology based on deep learning is carried out,and the experimental verification of deep learning target detection process and algorithm is carried out,which provides effective technical support for further vehicle applications.The main work of this article is as follows:First of all,this paper introduces the system structure and working principle of vehicle-mounted millimeter-wave radar.Starting from the radar equation,this paper analyzes the system structure and operating frequency of vehicle-mounted millimeter-wave radar,and conducts test verification platform selection and construction on this basis.These have laid the technical foundation for further technical research and experimental verification.Then,the thesis studies the signal model of linear frequency modulation continuous wave radar,at the same time studies the signal processing flow and processing model of the three-dimensional matching filter of distance,speed and angle,and the detection method and algorithm of target constant false alarm.Based on the test verification of the measured data,the performance of the above algorithm is verified.Finally,the thesis focuses on the performance loss caused by the model mismatch between traditional signal processing and target detection methods.It conducts research on target detection technology based on deep learning,trains and constructs a detection model based on convolutional neural networks.The model is verified by the measured data,which effectively solves the bottleneck problem of traditional multi-target detection,achieves a detection probability of better than 95%,and provides technical support for the future intelligent development of vehicle-mounted radar.