Research On 3D Building Semantic Modeling Based On Point Cloud And CityGML

With the continuous development of hardware equipment,3D laser scanning technology gradually plays a more important role in the construction of digital cities with its high precision,non-contact,automation and other characteristics.However,the problem of low automation caused by lack of semantic information and manual intervention has always been a challenging and urgent problem to be solved.As an open data model and supporting storage and exchange,CityGML is a three-dimensional model that can be used in a variety of applications.Compared with other data formats,CityGML has great advantages.CityGML has the characteristics of object semantics,geometry,topological expression consistency,etc.,and can realize multi-scale expression and refined modeling management in the field of 3D GIS.In this paper,by studying the characteristics of point cloud data,a semi-automatic method is used to obtain the semantic feature information of building point cloud data,and the CityGML standard three-dimensional building model is created.The main research contents are as follows:(1)Plane extraction of building point cloud data.Due to the difficulty of threshold determination in traditional area growth algorithms,random selection of seed points affects the effect of plane segmentation.In this paper,for a given threshold problem,an adaptive method is used to determine the search radius and select a robust seed point method to optimize buildings The effect of point cloud data plane extraction.(2)Extraction of semantic information of buildings,including:semantic information of facades,roofs,floors,doors and windows of buildings.The building plane is divided into roof,elevation and ground.In this paper,the semantic information of the extracted plane is distinguished by discriminating rules,and the boundary points of semantic information are extracted based on the method of neighborhood point analysis.Then,when there is a door and window structure in the facade,the boundary points are extracted,and the door and window information is extracted through the door and window discrimination rules.Finally,on this basis,a corner point is extracted based on the quadratic clustering algorithm to extract the corner points of the building's semantic feature information.(3)Point cloud data is converted to CityGML model.This paper designs a method to convert point cloud to CityGML model.In the CityGML data format,each polygon plane is formed by a closed ring,which is composed of a set of points connected to the beginning and end.In this paper,the extracted semantic corners of buildings are converted into the XML data format of CityGML,and all the semantic corners are combined to generate the final CityGML semantic model.