Research On Key Technologies And System For Intelligent Improvement Of Large Vehicles Based On Millimeter Wave Radar

In recent years,with the flourishing of the global automotive industry,and the prosperity of advanced vehicular radar sensors and artificial intelligence technology,automotive intelligent perception technology has achieved rapid development.However,this technology is still lacking for large vehicles.Due to the large volume of large vehicles,there is a blind spot in the field of vision behind the body,so there are certain risks in the reversing process.Moreover,the rapid development of the logistics and transportation industry also puts forward higher requirements for the intelligent construction of large vehicle fleets.However,large vehicles cannot effectively identify the driver’s identity,which limits the development of intelligent management technology.To improve safety during the reversing process and promote intelligent management,this study designs a vehicular radar perception system suitable for large vehicles,combining vehicular radar sensors and the latest target detection and recognition technology.The main contents of this research are as follows.(1)To improve the safety of large vehicles during the reversing process,this research designs a blind spot detection scheme based on the fusion of millimeter wave radar and ultrasonic radar.First,this research proposes a millimeter wave radar target detection process based on point cloud clustering,target association,and target tracking.After that,given the own characteristics of the millimeter wave radar point cloud,this research proposes a PV-DBSCAN clustering algorithm based on polar coordinates and voxelization,which can not only improve the accuracy of cloud clustering but also promote the efficiency of computation.Besides,to meet the actual application needs,the decision-level fusion strategy of millimeter wave radar and ultrasonic radar is designed to realize the complementary advantages of both.(2)To promote the intelligent management of large vehicles,this study designs a gait recognition method for drivers based on millimeter wave radar.In this study,the point flow information is proposed,which can represent the dynamic characteristics between frames of the radar point cloud.Besides,this study proposes HDNet for gait recognition.This network includes a dynamic frame sampling module to promote the efficiency of computation without deteriorating performance noticeably.This network can achieve an accuracy of 96.52% on the public gait recognition dataset,which is superior to other comparison methods,and still has a high recognition accuracy when the sampling ratio of point cloud frames is low.(3)Based on the above two parts,this research designs a vehicular radar perception system.This system includes the blind spot detection subsystem installed at the rear of the large vehicle and the driver gait recognition subsystem installed at the door position.First,this study introduces the overall framework of the system and designs a complete communication process and visualization scheme.Secondly,this study analyzes the feasibility of driver gait recognition in practical application scenarios and designs a complete recognition process.The recognition accuracy achieves 94.32% on the dataset collected in the practical scenarios,and the open set recognition AUROC reaches 81.73%.Finally,the function and performance of the vehicular radar perception system are tested in this study,which verifies that the system can meet the application requirements of the actual application scenarios.