Study On Building Facade Modeling Of Combining Airborne-vehicle LiDAR Point Cloud

Building is the main body of the three-dimensional scene of a city,and building three-dimensional modeling is one of the core contents of building a digital city and a smart city.With the rapid development of China’s urbanization and the arrival of the era of big data,the requirement of building three-dimensional modeling is more authenticity,automation,refinement and rapidity.Airborne LiDAR and vehicle-borne LiDAR are two main technical means to obtain 3D information of buildings in large scenes,which have complementary advantages.Vehicle-borne LiDAR system can quickly acquire the point cloud data of building facade,but due to the influence of occlusion and inaccessibility of ground objects,there are serious data incompleteness in the obtained point cloud of building facade,which brings troubles to building facade modeling.Airborne LiDAR system acquires the roof cloud data of buildings,and has limited ability to acquire the elevation information of buildings.Facade is the main bearing part of building details.Due to the incompleteness of point cloud data and other constraints,it is difficult to build an accurate and complete building facade model only by using the vehicle-borne LiDAR system.From the spatial distribution of urban buildings and the structural characteristics of buildings themselves,there are many conjugate feature information and spatial semantic association information between airborne LiDAR and vehicle-borne LiDAR.Airborne LiDAR can provide abundant information and semantic support for building elevation modeling.Based on between the spatial topological relationship of building facets and the geometric and semantic description of the details of building facade,this paper presents a method of building facade model modeling combined with airbornevehicle LiDAR point cloud.The main research work is as follows:1 Aiming at the research work of data preprocessing and building model construction based on airborne LiDAR and vehicle-borne LiDAR data sources at home and abroad,the problems and shortcomings of building model construction with two kinds of data sources are deeply discussed and analyzed,the advantages of building elevation model based on combination of airborne and vehicle-borne LiDAR are given.2 The preprocessing methods of airborne and vehicle-borne LiDAR point cloud data are optimized.The CSF filtering algorithm is used to separate ground points from non-ground points in airborne point cloud data,and the inverse distance weighted difference algorithm is used to construct a regular DEM grid for the separated ground points.The DEM is used as elevation datum to extract buildings by setting reasonable elevation threshold.The data of LiDAR point cloud in vehicle is pre-processed by data block method,and noise points around buildings are filtered by SOR filtering algorithm.3 A set of modeling method of building elevation wireframe model based on airborne-vehicle LiDAR point cloud is proposed.Firstly,two pre-processed data sources are registered,and an improved eight-neighborhood clustering algorithm is used to cluster and segment the registered roof cloud.Then,the contour points of the roof are detected by the dynamic elliptic convex hull algorithm and vectorized based on the principle of grouping orthogonality.A buffer is constructed to segment the elevation of the building.The improved RANSAC algorithm is used to detect the main wall of the building,and the wiring frame model of the building elevation is constructed based on the topological relationship of the building facets.4 The modeling method of building elevation refinement model is deeply studied.For the detected point cloud on the main wall,firstly,the windows and doors in the facade are clustered based on the idea of holes,and then the dynamic elliptic convex hull algorithm is used to detect the boundary points of windows and doors.Because of the influence of the quality of point cloud,there will be some windows whose boundary points can not be detected.To solve this problem,this paper restores the missing windows according to the similarity between the structural units of windows.The boundary points of windows and doors are segmented by RANSAC algorithm and vectorized by robust total least squares algorithm.Finally,the three-dimensional model of windows and doors is constructed by combining the geometric characteristics and semantic description of elevation structure units.5 Experimental verification and reliability analysis based on the proposed algorithm.Selecting the airborne and vehicle-borne LiDAR data of the experimental area,the experiments of roof cloud preprocessing,building elevation segmentation,main wall detection,structural unit clustering,outline point detection,missing window repair,structural unit boundary segmentation vectorization and fine building elevation model construction are carried out respectively.The results are analyzed qualitatively and quantitatively.The algorithm in this paper can accurately construct the elevation model of fine buildings.