Typical Ground Features Extraction And Scene Virtual Analysis Of Road Scene Based On Vehicle-mounted LiDAR Technology

There are many kinds and quantities of ancillary features in urban road scene,so the inventory and management of road assets is a difficult work.Acquisition of location and quantity information of typical ground objects in urban roads by manual measurement is not only a heavy workload but also a potential safety risk.The development of Vehicleborne Light Detection and Ranging(Vehicle-borne LiDAR)technology provides a new solution for the three-dimensional information of road scenes.Vehicle-mounted LiDAR measurement system can quickly and efficiently acquire massive point cloud data of road scenes.How to scientifically and effectively manage,analyze and apply these information is of great significance to municipal management and is also a research hotspot of many scholars at home and abroad.This paper takes typical ground features in urban road scene as the research object,and carries out a series of related researches and explorations around ground features classification,modeling,scene simulation and related applications.The main work contents are as follows:(1)The application and development of current vehicle-mounted LiDAR technology in road scenes are deeply analyzed.This paper reviews the development history of vehicle-mounted LiDAR technology,expounds the domestic and foreign research status of classification extraction,model building and road scene virtual reality representation based on vehicle-mounted LiDAR point cloud data,analyzes the shortcomings of the current research,and puts forward detailed solutions on this basis.(2)The on-board LiDAR data preprocessing process is refined and standardized.The composition and working principle of vehicle-mounted LiDAR measurement system are introduced,and the pre-processing methods and processes of vehicle-mounted LiDAR data are refined and standardized for the research objective,and the pre-processing works such as data calculation,tailoring,denoising and partitioning are introduced in detail.(3)An improved Bagging ensemble learning classification method based on multiple voting is proposed,which makes full use of the complementary information among different classification algorithms to realize the classification of typical ground objects in road scenes.To remove the ground point to clustering segmentation of point cloud data,building all the eigenvalues of the clustering unit,then select more than one can effectively distinguish between different types of features of the characteristic value of characteristic vector,use a variety of machine learning algorithms of point cloud data,obtain the prior knowledge of the algorithm,finally,the classification algorithm is integrated with better effect.Experiments show that the method is effective and there are complementary information among different algorithms.(4)Set up a virtual road scene.According to different features of the characteristics of point cloud data,using different modeling methods,respectively,based on 3 d visualization software Lumion platform to realize the road scene virtual expression,and road scene lighting,for example,has been applied to build the road scene,this study is expected to be quantified for road lighting and fine management to provide strong technical support.