Research On 3D Modeling Method Of Buildings Based On The Airborne And Vehicle-borne LiDAR Data

With the development of mobile measurement technology,urban spatial information acquisition has been greatly improved.LIDAR(Light Detection And Ranging)is an emerging spatial data acquisition technology.Both airborne and vehicle-borne LiDAR can quickly collect the 3D urban point cloud data of a large area,which have become important sources for 3D city modeling.Because the data collection methods of vehicle-borne and airborne LiDAR are different,the two data sources have their advantages and disadvantages during modeling,which can mainly be shown as the following two aspects: 1)the models constructed from the airborne LiDAR point cloud can express the roof structure of buildings better,but lacking the expression for the building wall structure;2)the models built from the vehicle-borne LiDAR point cloud can express the building walls better,but the expression for the roof structure is lacking;meanwhile,limited to the collection conditions and tree occlusion,the building wall models constructed are often incomplete or even missing.Therefore,in order to meet the requirements for precising building modeling,this paper conducts the research on the 3D building modeling method based on the fusion data of airborne LiDAR and vehicle-borne Li DAR.This paper analyzes the current research on the automatic building extraction and reconstruction from airborne and vehicle-borne Li DAR point clouds.On the basis of data fusion of the airborne and vehicle-borne LiDAR point cloud,the paper focuses on the automatic building extraction and 3D building reconstruction.The specific contents are as follows:1.Do an analysis on the current researches of the automatic building extraction and reconstruction based on airborne and vehcile-borne LiDAR and fusion data of the two.Respectively describe the data characteristics of the airborne and vehcile-borne LiDAR and the data fusion methods.2.On the basis of data fusion,analyze the automatic building extraction from airborne LiDAR point cloud,using a stepwise filtering strategy to extract the building point cloud;An eight neighborhood connectivity analysis methods based on grid density is adopted to achieve the clustering for every single one building.For eachbuilding,an improved local convex hull algorithm is used for the building contour extraction,and outline simplifiction and regularization are carried out.3.Based on the extracted outlines,the Li DAR point cloud of building walls can be automatically segmented,and the adjacent building similarity is calculated through the contour shape analysis.According to the building similarity and segmented wall point cloud,the same parts of the similar building are used for hole filling and missing wall repairing.4.An improved RANSAC algorithm is used for roof patch segmentation,and then the airborne building modelings are constructed.Through wall feature extraction,achieving the building wall and window reconstruction;Based on the building top and wall models,the fine fusion building model is constructed.5.An experimental area including airborne LiDAR and vehcile LiDAR point cloud data is selected to verifiy the data fusion,building extraction,contour extraction,wall segmentation,roof and wall models constructing,building model fusion,etc.Then the qualitative and quantitative analysis are used to evaluate the building reconstruction effect,and the results show that this method can effectively build finer3 D building models.