Research On AR-HUD Navigation Induction Model And Algorithm Based On Multi-source Information Fusion

Vehicle navigation function is very important in the process of driving.Currently,vehicle navigation software is divided into vehicle-mounted navigation and mobile navigation,both of which will more or less affect the driver’s attention.To solve this problem,the augmented reality head-up display system(AR-HUD),which combines AR and HUD,integrates the map navigation information with the real lane information obtained by environment perception,and projects it onto the front windshield in the form of virtual image,superimposed on the real road environment,so as to play a more accurate and intuitive navigation induction effect.Therefore,the navigation technology of vehiclemounted AR-HUD has become the inevitable trend of the next generation of navigation.This dissertation studies the navigation induction model and algorithm in the navigation function of AR-HUD system,which is based on multi-source information fusion.The primary research tasks are described below:1.In view of the current navigation information from a single source,there are problems such as incomplete vehicle-mounted navigation information,inconvenient update,insufficient navigation and positioning accuracy,erroneous inspection exists in machine visual detection.In this dissertation,an AR-HUD navigation system architecture based on multi-source information is designed by combining input information such as ADAS visual perception system,Android mobile phone navigation system,vehiclemounted navigation system,car body system and in-car camera.2.In response to the acquisition of multi-source information of AR-HUD system and data matching and navigation accuracy problems in the fusion.This dissertation designs a data acquisition and analysis method of Auto Navi 2D electronic map,and a lane line information detection algorithm,which is based on monocular vision for dynamic region of interest.The experiment proved that it can obtain certain navigation map information,lane line parameters and road information in the real scene such as trend.Further,designed a logic matching fusion algorithm based on voting method.The voting method was used to fuse the navigation map information and ADAS visual information in realtime.The results show that the information detection rate after fusion is higher than that of single identification method,and the driving area in front of the vehicle can be determined by using more reliable lane information identification results and logical matching.3.Constructed the navigation induction model and algorithm under a variety of typical scenes,including straight line,turn left,turn right,change lane,slope,etc.Each module of AR-HUD system software is designed and implemented,and completed the development of the AR-HUD prototype system,which has navigation guidance functions of various scenes based on the Linux vehicle-mounted platform.Finally,the simulation test and actual vehicle debugging of the corresponding scene are carried out to provide intuitive and accurate navigation instructions and give users an immersive driving experience.