| The extraction of indoor navigation network is the foundation of indoor navigation work,and plays an important role in the subsequent real-time positioning and navigation.The traditional navigation network construction based on the two-dimensional plane of the building can not accurately express the three-dimensional geometric environment information inside the building.IFC and CityGML data source is difficult to obtain and update.Therefore,as a data source that has been widely used in recent years,3D point cloud data which is easy to obtain provides a new solution for the construction of indoor navigation networks.Considering the service of indoor navigation,this paper proposes a method of generating indoor navigation network based on point cloud by using the geometric and statistical characteristics of the whole building point cloud.Aiming at the problem that the semantic information and topology information in the scattered point cloud are difficult to extract,the method of extracting the topological relationship from top to bottom is introduced.Especially for the room in the navigation element,the idea of refining the space separation is proposed.The point cloud density binary map is used for region growth to obtain the room expression that retains the indoor details.while using the corresponding positions of the two wall surfaces on one wall,point cloud density information comparison avoids errors caused by obstacles in door and window extraction,and improves the accuracy of topological relationship definition.The experimental results show that the indoor three-dimensional navigation network extracted by the method described in this paper solves the problems of insufficient semantic and topology information in the traditional method,and can provide a model basis for indoor real-time positioning and navigation.The main research contents of this paper include:(1)Building indoor navigation elements extraction and semantic annotation: Based on the in-depth analysis of the existing indoor navigation element algorithm,the method of analyzing the point cloud in the large scene indoors is studied,and the navigation elements of each level are gradually extracted.Including floors,walls,rooms,doors and windows,semantically annotating them,wherein the floor segmentation is performed by using the peaks in the point cloud density histogram;the room extraction is to generate a binary map by tiling a single floor,two The value map uses the region growing algorithm to obtain the corners of the room to form the geometric and topological representation of the room;the door and window extraction is obtained by comparing the point cloud density values of two walls on one wall.The navigation elements of each navigation element obtained by the data model are designed.(2)Topological relationship definition of the extracted navigation elements in the process of step-by-step extraction from top to bottom,mainly including topology inclusion relationship(between floors and rooms),topological adjacency(such as between floors,rooms)And topological connectivity(such as between floors and rooms).When the parsing is performed hierarchically,the relationship between the subspace and the upper space is automatically defined.The subspace separated by the shared space is defined as an adjacency,and the two subspaces with open regions on the shared spatial partition are defined asconnected relationships.(3)Automatically generate an indoor navigation network including navigation element nodes and topological relationships according to the extracted navigation elements and their topological relationships,realize navigation network storage and expression?... |
PDF Full Text Request