Research On Road Target Detection Method Based On Millimeter Wave Radar And Vision Fusio

In recent years,our transportation industry has developed rapidly,the development level of transportation has been continuously improved,and the construction and operation of infrastructure has achieved remarkable results.Data shows that there are about 250,000 intersections and 4 million kilometers of road sections in China.Highway infrastructure has greatly facilitated people’s travel life,enhanced road transport service capacity,and promoted social and economic development.However,with the continuous expansion of the urban scale,the urban population and the number of vehicles increase sharply,and the traffic congestion coefficient keeps rising.At the same time,all kinds of traffic violations and traffic accidents occur frequently,especially at the intersections of urban trunk roads and mixed roads,which further aggravate the congestion of urban roads,reduce the safety of residents’ transportation,and bring new challenges to the development of urban transportation.It is imperative to realize the intelligent construction of the city through new ideas and new technologies to solve the two major pain points of traffic congestion and accidents.In order to solve the above problems,this paper studies and compares the defects of current road target detection technology.After consulting the relevant data,the road target detection method based on sensor fusion is proposed,so as to realize the function of monitoring the intersection traffic and facilitate the traffic management system to count the intersection flow later.The main work of this paper is as follows:First of all,refer to the literature related to road target detection,analyze the current technology,summarize the shortcomings of the existing technology,and finally determine the road target detection method with sensor fusion as the core.Secondly,ARS408-21 radar produced by Continental Germany and IMX415 vision sensor produced by SONY were selected as the two sensors for fusion.Analyze the data of two sensors,including: analyze the original radar data,eliminate the data that does not meet the requirements,complete the road target screening;The visual sensor was used to collect data to make data set training network,and the original YOLOv5 network was improved to make the improved visual network model fit the complex traffic scene,and the final recognition accuracy reached 91%.Thirdly,after the data processing of the two sensors is completed,the self-made checkerboard is used to calibrate the visual sensor by Zhang Zhengyou calibration method.Millimeter wave radar and vision sensor are fused by coordinate conversion and thread synchronization.The Hungarian algorithm is used to judge the consistency of radar target and visual target.Finally,relying on the research group to build a test platform,debug equipment at different traffic intersections and conduct experiments.By comparing the detection results of the vision sensor and the fusion system respectively,the effectiveness of the fusion model in the paper is proved,and the advantages of the multi-sensor fusion model in the field of road target detection are also proved.