Research On Road Element Extraction Method Of Vehicle-borne Laser Scanning Data

As a new type of mobile measurement technology,vehicle-borne mobile measurement system has obvious advantages compared with traditional measurement technology in fast acquisition of spatial information.It can not only quickly collect high-precision threedimensional coordinates of the surface of the object around the vehicle trajectory,but also obtain its texture,reflection intensity and other information.It has been widely used in the fields of unmanned driving,road maintenance,urban planning.The classification and extraction of vehicle-borne LiDAR point cloud data is the key to many applications.However,due to the problems of massive volume,uneven density distribution,and occlusion and self-occlusion between targets in the scene,the vehicle-borne laser scanning data has low practical utilization efficiency.In view of the above problems,this paper studies and summarizes the classification and extraction algorithms of target ground objects in vehicle-borne laser point cloud data,deeply analyzes the geometric information and spatial distribution characteristics of various ground objects in the road environment,and designs the extraction algorithm of road boundary,road surface,road marking line and street lamp.In this paper,the road element point cloud extraction algorithm designed mainly includes the following three parts:(1)An improved Euclidean clustering algorithm for road boundary and road surface extraction is proposed.The algorithm consists of four steps: the first step is to filter the ground points by adaptive filtering based on scanning lines;the second step is to segment the filtered point clouds by using the Euclidean clustering algorithm with smoothness constraints;the third step is to optimize and track the point clouds for each clustering,filter out the vehicles,trees,street lights and other objects,and then extract them completely.In the fourth step,the road boundary is taken as a constraint condition to extract the point cloud of the road surface.(2)Road marking line extraction and classification algorithm based on contour analysis.The algorithm firstly preprocesses the point cloud data,and calculates the minimum bounding box of the point cloud after removing the outliers.Then,the resolution of the intensity feature image is calculated according to the bounding box size and the grid interval,and the feature image is generated through projection;The Canny operator is used to detect the edges,and the contours are assembled by connectivity analysis.Then analyses the size and shape of each contour.After filtering the outline of the unlabeled line,classifying the road marking line,restoring the classified marking line to point cloud according to the index;finally,the point cloud of the marking line is refined by using the Gauss mixture model to obtain the accurate road marking line.(3)Street lamp extraction algorithm based on the two-three dimensional morphological features.According to the morphological characteristics of street lamp,the method detects the rod-shaped object and the lamp head separately to extract the lamp point cloud.Firstly,the road boundary and the road point cloud in the vehicle laser point cloud data are filtered out;then the street lamp candidate point cloud is extracted according to the two-three dimensional shape feature of the rod-shaped object;finally,based on the analysis of the spatial morphological differences of street lamp head,the traffic sign sign surface and the tree canopy,the coverage analysis and projection area analysis of the candidate point cloud of street lamps were carried out to extract the complete point cloud of street lamps.In this paper,three sets of vehicle-borne point cloud data under different scenarios are experimented to realize automatic extraction of road boundary,road surface,road marking line and street lamp point cloud,which verifies the applicability and feasibility of the extraction and classification algorithm in this paper.