Research On Vehicle Detection Method Based On Radar And Vision Fusion

With the rapid development of computer technology and artificial intelligence,China regards the development of intelligent connected vehicles as a new development strategy,and the intelligence of vehicles has become a new research direction in the field of vehicle engineering.For intelligent vehicle,its environment perception system is an indispensable key technology to realize automatic driving.This paper proposes a vehicle detection method based on the fusion of millimeter wave radar and vision sensor,which can provide reliable data for the planning and decision-making of intelligent vehicles.Firstly,this paper analyzes the functional characteristics and working principle of millimeter wave radar,and completes the selection of millimeter wave radar.Basing on the selected millimeter wave radar,this paper analyzes the received data of millimeter wave radar according to the communication protocol,and obtains the data of obstacles in front of the vehicle.On this basis,this paper completes the forward vehicle detection based on millimeter wave radar through screening and visualization of effective targets detected by millimeter wave radar.Next,based on the analysis of the traditional visual target detection method and target detection method based on the deep learning,this paper analyzes the principle of the convolution of the neural network based on deep learning,sets up the visual target detection model based on YOLOv4 target detection algorithm.Training target detection model by using traffic scene dataset,verifying and testing the model results after training,this paper completes the forward vehicle detection based on vision sensors.Then,this paper analyzes the principle and hierarchy of multi-sensor fusion,and through the transformation of the coordinates of the millimeter wave radar data and visual sensor data completes the space of sensor fusion,through the adjustment of two sensors’ sampling time completes the time of sensor fusion,and using the data association algorithm based on global nearest neighbor realizes the sensor data association.A vehicle detection system based on the fusion of millimeter wave radar and vision sensor is established.Finally,by establishing an experimental platform equipped with millimeter wave radar and visual sensor,this paper carries out vehicle forward target detection experiments in different traffic scenes.By comparing the target detection results of a single sensor system and a sensor fusion system,this paper verifies the effectiveness of vehicle detection method based on the fusion of millimeter wave radar and vision sensor researched in this paper and proves its advantage in the field of vehicle detection.