Research On Vehicle Detection And Tracking Algorithm Based On Millimeter Wave Radar And Deep Vision Information

With the increase in the number of cars and the increasing complexity of the traffic environment,the problem of traffic safety accidents caused by cars has become increasingly prominent.It is urgent to reduce the incidence of traffic accidents and improve the safety of cars through driving assistance systems and advanced automatic driving technologies.Sex and efficiency.Among them,environment perception technology is a key technology in autonomous vehicles.As the eyes of the car,it always recognizes the information about the surrounding environment of the vehicle,and transmits the recognition results to the decision-making and planning layer of the car for further processing.In this paper,vehicles on urban highways are used as the research object,and vehicle detection and tracking are carried out based on millimeter wave radar and deep vision information fusion algorithm.The main research contents are as follows:(1)Millimeter wave radar detection and tracking based on adaptive extended Kalman filter.First,analyze the working principle of the millimeter-wave radar and analyze the detection data to extract the original data of the millimeter-wave radar detection.Then,the effective target of the radar is selected and the validity is verified.Finally,an adaptive extended Kalman filter multi-target tracking method is proposed,which combines the joint probabilistic data association algorithm to associate multi-target data and track management.The research results show that the algorithm can continuously and effectively track vehicle targets.(2)Construct a convolutional neural network structure and train annotated data sets for vehicle detection and tracking.First of all,to solve the problem of excessive training parameters of traditional SSD target detection algorithms and difficulty in training convergence,the SSD+Mobile Net-V2 convolutional neural network is constructed.Secondly,manually label the vehicle data set label,construct a complete vehicle data set and input it to the convolutional neural network for training.Finally,a multi-target tracking algorithm based on visual images is proposed.The research results show that the algorithm can improve the stability of visual inspection and reduce the missed and false detections of vehicles.(3)The fusion of millimeter wave radar and depth visual information.First,align the space coordinate system and the time coordinate system of the monocular camera and the millimeter wave radar.Secondly,the radar detection target point is projected to the visual image to complete the generation of the radar area of interest.Finally,a decision-level fusion strategy is proposed to output the fusion result of radar and vision detection.The research results show that the fusion algorithm has better detection accuracy.Based on the python language,this paper uses the pycharm compiler and the pytorch neural network framework to build a vehicle fusion detection algorithm,and collects highway data to verify the algorithm.The results show that the algorithm proposed in this paper can effectively detect and track multiple targets in front of the vehicle,and has good environmental adaptability and real-time performance.