| With the acceleration of urbanization,complex indoor scenes in buildings have become a cognitive obstacle for people to find their way.For example,in the emergency evacuation,complex indoor scenes will greatly increase the time cost of personnel evacuation and cause losses to people’s life and property safety.Therefore,in the face of complicated indoor scenes,it is necessary to analyze the spatial structure of indoor scenes and build an easily identifiable indoor navigation network model so as to quickly plan the optimal path,so as to protect people’s life and property safety.This paper optimizes the algorithm of indoor scene model reconstruction and indoor navigation network model,and integrates indoor scene model and indoor navigation network to propose a navigation network model construction method based on 3D point cloud under indoor scene.The main idea of this paper can be divided into the grid expression of indoor scenes,the generation of indoor space topological relations and the construction of navigation network model under indoor scenes,so as to form the analysis and understanding from the local to the whole.The main research work of this paper is as follows:(1)Grid expression of indoor scenes.In order to avoid the existence of narrow triangles in indoor scene model,Incremental insertion algorithm meshing algorithm based on KD tree is proposed.Firstly,incomplete segmentation KD tree index is constructed for point cloud data.Secondly,control points(non-leaf nodes after KD tree division)are introduced to adjust the insertion sequence of point cloud data.Finally,Incremental insertion algorithm is used for triangular mesh reconstruction to realize grid representation of indoor scenes.(2)The generation of topological relation of indoor space.The topological relation of indoor space is the data support of the indoor navigation network model.The location information and structural relation of each indoor space unit can be determined through the connectivity relation,the adjacency relation and the inclusion relation.Aiming at the generation of topological relations of indoor space,Point Net++ was introduced to realize the extraction of indoor space units,and NRG model was generated on the basis of IndoorGML to construct the connectivity,adjacency and inclusion relations among indoor space units,and to realize their visual expression.(3)Construction of navigation network model in indoor scenarios.Aiming at the disadvantage of unreadable visual representation of traditional indoor navigation networks,a navigation network model construction method in indoor scenes was proposed.The indoor navigation network model can be constructed based on the topological relationship and spatial database of indoor space,and the indoor navigation network model can be embedded into the indoor scene model to improve people’s cognitive ability of complex indoor space structure.In this paper,3D point cloud data of indoor buildings were used as data sources to design experiments to verify the feasibility of grid representation of indoor scenes,topological relationships of indoor Spaces and navigation network models under indoor scenes.The proposed optimization algorithm of indoor scene model is verified from the aspects of time efficiency and model quality,and the experiment shows that the optimization algorithm is superior to other comparison algorithms.Taking Bunny’s triangular mesh reconstruction as an example,the optimization algorithm has increased the time cost by 13% compared with BRIO and 8% compared with KD tree meshing algorithm.As for the generation of topological relations in indoor space,m Io U of Point Net++ semantic segmentation was 0.5243,and the average classification accuracy reached 83%,which was in line with the expected results of the experiment.Moreover,the topology of indoor space generated based on NRG is clear,and the visualization of the topology of indoor space is realized.In the construction of navigation network model in indoor scene,the model integrated by indoor navigation network model and indoor scene model not only maintains the connectivity of navigation network but also provides external reference for navigation network.Finally,the correctness of spatial topology and the feasibility of navigation network are verified by using Dijkstra algorithm for indoor navigation network model. |
