Cognitive Map-Oriented Intelligent Vehicle Localization System And Application

The positioning of conventional intelligent driving vehicles relies on satellite positioning,laser radar or image matching.The positioning accuracy of ordinary GPS devices is poor(2-5m).High-precision GPS devices have high hardware input costs.Navigation systems based on satellite positioning exist.The problem of signal dependence,that is,when the signal is blocked on the road,can not complete the system positioning;second Lidar positioning can achieve high-precision positioning through the point cloud registration,but its high cost,large amount of data,large calculations and other shortcomings make Lidar is not suitable for the next generation of smart car development;image matching positioning relies on image databases,and the storage of image databases requires huge storage space.The search and matching process of images in databases is computationally intensive and does not lend itself to large-scale image building.Positioning.In view of the above problems,this paper aims to complete a smart car positioning system: the map data is small and the amount of calculation is small.This dissertation focuses on the research of vehicle location technology based on cognitive maps.The main research work is as follows:Monocular vision based cognitive map creation.The members of cognitive maps in this thesis are traffic signposts.For the disadvantages of high cost and large amount of laser sensors,and the need for manual marking of road signs,this paper proposes to create maps visually and adapt to the environment.Strong,high-precision maps,and a small amount of storage required,can meet the requirements of the smart car navigation and positioning of the map.Cognitive Map Based Intelligent Vehicle Positioning Method.This paper uses pure monocular vision to complete the global positioning of the smart car.The realization of this method calculates the position and orientation of the smart car by retrieving the threedimensional coordinate information of the landmarks in the current frame and the imaging position of the landmarks in the two-dimensional image..This part is divided into two research directions: self-positioning of smart cars based on three traffic signs,and selfpositioning of smart cars based on individual traffic signs.Smart car positioning system.The research of this paper is oriented to the practical application,aiming at the requirement of platform configuration for the paper research to build a smart vehicle experimental platform.Cognitive map creation and smart car location module based on cognitive maps are integrated to complete the automated creation of offline cognitive maps in campus scenes and map-based smart car location.The goal of the paper's smart car positioning system is to be able to be widely used in the next generation of smart car positioning,which can greatly reduce the system cost and reduce the amount of map data,providing the technical foundation for the next generation of smart cars to truly enter society.