OpenStreetMap-Assisted Extraction Of Road Geometric Feature From Mobile Laser Scanning Point Clouds

The mobile laser scanning system is equipped with a LiDAR system on a moving vehicle.It can acquire high-precision three-dimensional space information of roads and surrounding environment quickly,continuously,and at low cost,which has a wide range of applications,such as urban planning,intelligent transportation,and road environment monitoring and many other fields.The road centerline and boundary is the basic information of the road and is a prerequisite for the geometrical estimation of the road and the detection of the ground objects around the road.Therefore,it is very meaningful to extract the high-precision 3D road centerline and boundary from the vehicle-borne LiDAR point cloud.However,the vehicle-borne LiDAR point cloud has the characteristics of high density and non-uniformity,large amount of data,incomplete data due to occlusion,and complex scenes,which makes it difficult to extract road feature from vehicle-borne LiDAR point cloud.Based on the results of previous studies,the paper focuses on the research topic of the road centerline and boundary extraction of vehicle-borne LiDAR point clouds,and proposes a method based on active contour model with the assistance of OpenStreetMap(OSM)data.our method obtained complete three-dimensional vector centerline and boundary information with attribute data.The research significance of this article is then explained with an actual application.The main research content of this article is(1)The background and significance of the study are expounded.Under the background of the development of LiDAR technology,it is very important to extract the road information from the vehicle laser point cloud.(2)Domestic and foreign literature on the extraction of vehicle-borne laser point cloud roads is studied,and the existing algorithms is classified into four major categories based on scan lines,point cloud feature map,clustering,and machine learning.Some representative algorithms of each catagory is described in detail.Then the advantages and limitations of each method is summarized.We analysis the advantages and disadvantages of vehicle-borne laser point cloud data.The shortcomings of the data illustrate the difficulties of extracting road targets from the vehicle-borne laser point cloud.(3)The literature about applying Snake model to road extraction is researched,and the feasibility of Snake for road extraction of vehicle-mounted laser point cloud is illustrated.The basic idea of the traditional Snake model,the mathematical expression and the convergence strategy of the energy function are introduced.Then the Ribbon Snake model which is improved in the traditional Snake model to make it suitable for road extraction is introduced.Then the Snake model is compared to other common edge detection algorithm to explain the advantages of using Snake model.We propose an improved model for the Ribbon Snake model to make it suitable for road extraction of vehicle-borne LiDAR point cloud.(4)OSM data which is a typical kind of Volunteer Geographic Information is introduced including the data structure,the tagged attribute information.The advantages and disadvantages of OSM data is elaborated.Vehicle-borne laser scanning data has high accuracy,three-dimensional coordinates,but has no semantic information.OSM data has rich semantic information,but its accuracy is not high.The OSM data and vehicle-borne laser point cloud data can complement each other.This article uses the snake model to intelligently correlate the two data.Snake is initially the location of the OSM and ends at the road boundary of the point cloud.(5)Vectorization and three-dimensionalization are performed for the grid boundaries on the feature map.For the pixels which has no point in caused by occlusion,the range of the pixel is interpolated using morphology to obtain accurate elevation information.The semantic information and attribute information of OSM are assigned to a three-dimensional road boundary to form a complete three-dimensional road data with both accurate geometric information and attribute information.According to the point cloud feature at curb ramp,the exact position of the ramp on the boundary is determined so as to make reasonable road planning.