Forward Vehicle Detection Based On Millimeter Wave Radar And Visual Information Fusion

With the rapid development of Chinese economy,automobiles are gradually entering millions of households as a highly convenient means of transportation,and with it,the increasing number of traffic accidents.In order to improve the safety of driving,the Advanced Driver Assistance System(ADAS)has been created,the key of which is the use of various on-board sensors to perceive the driving environment,thus providing the basis for the driver and the vehicle system to make appropriate decisions and control.Among them,Vehicle detection technology is one of the important technologies in environmental perception.Millimeter wave radar and camera are two common sensors used in ADAS systems for perceiving the environment.Millimeter wave radar has good environmental adaptability,strong penetration ability,and can accurately detect the speed,azimuth,distance and so on.However,it is limited by its working principle,high detection noise,excessive clutter interference data,and inability to obtain target geometry and category information.Vision sensors,with their low cost and rich amount of data information that can be acquired,are widely used for target identification and classification,but they are susceptible to environmental factors such as light and weather.Therefore,if millimeter wave radar information can be fused with visual information,the advantages of both can be combined to achieve the purpose of improving the accuracy and real-time of vehicle detection.The main research contents of this paper is as follows:1)Radar data pre-processing to achieve effective target determination.Based on the characteristic analysis of the target information detected by millimeter wave radar,the interference targets are classified,and then the corresponding methods are taken to filter out these interference targets.2)Car detection based on car image features.A method is used for car detection,which is based on a combination of shadow features under the car and symmetry features of the car.First,the image is pre-processed with grayscale and image segmentation,then the morphological operation is performed to determine the region of interest(ROI,Region of Interest),and finally the car symmetry feature is used to verify the existence of the car3)Car detection based on machine vision.For the shortage of vehicle detection based on image features,an Adaboost algorithm combined with Haar-like features is selected to train the Haar classifier for vehicle detection,and then the algorithm is implemented using Opencv in combination with VS2019.4)To perform the fusion of millimeter wave radar and visual information.The region of interest is generated by projecting the radar information into the image through spatial coordinate system conversion and sensor time synchronization,and the vehicle detection is synchronized by combining the image based features with machine vision,and then the image ROI generated by both is correlated with the information,so as to realize the vehicle detection based on the fusion of millimeter wave radar and vision information.It is experimentally verified that the method proposed in this paper can effectively detect the vehicle in front with high real-time and accuracy under good lighting conditions and clear field of view,which provides key information for ADAS system to realize vehicle road traffic environment perception.