Research On Regular Building Model Reconstruction And Optimization Method Based On LiDAR Point Cloud

Three-dimensional reconstruction of buildings is a hot spot in the field of spatial information technology in recent years.A perfect 3D model of buildings is an important precondition and basis for the construction and development of numerous virtual platforms such as "digital city" and "digital earth".In order to express the building better and to enhance the cognitive effect of the building simulation to obtain the complete details of the appearance and structure of the building,a three-dimensional expression of the geometric structure of the building should be performed to construct a realistic three-dimensional model,so as to overcome the shortcoming of two-dimensional expression that visual perception is not intuitive.In recent years,with the development of LiDAR laser point cloud detection technology,large-scale building point cloud data can be easily obtained by various ways,laying the foundation for the study of building point cloud reconstruction.In this paper,aiming at the existing techniques for 3D mesh model reconstruction about existing building laser point cloud data have some questions that the accuracy of the reconstruction model is not high,the modeling process is complex and time-consuming and so on,adopting the two-step method to design and construct a model reconstruction optimization algorithm based on semantic rules.The algorithm realizes the mesh model reconstruction of the building point cloud data,and optimizing the output mesh model on the basis of maintaining the overall geometric structure of the model.First of all,using the measurement parameters and other information of LiDAR laser scanning system,the initial dense point cloud data is filtered and denoised by filters.Based on the complete geometric details of the point cloud data,the subsequent calculation pressure is reduced,and the filtered and denoised point cloud data is spliced by a registration algorithmto obtain a complete data set of a large-scale building scene,then the target point cloud is segmented from the entire point cloud data by using manual interaction method,so as to extracting the point cloud data of the target building accurately,then reconstruct the after preprocessing Li DAR point cloud data of the building using GreedyProjection to generate an initial building model of triangular mesh.Next,based on the mesh reconstruction of point cloud data,the initially constructed triangular mesh model of building is optimized,optimization operations include regular wall optimization and junctional zone optimization.Aiming at the regular wall of a building,firstly,we use interactive mode to set a semantic constraint surface determined by three correct points for each regular wall of the initial triangular mesh mode,which is simply referred to as a regular surface,do intersection calculation between the regular surface and the triangular mesh of the corresponding wall of building model,then through the projection calculation,the triangular facets intersecting with the regular face are projected onto the regular face,so as to complete the optimization of the uneven wall based on the regular face.Aiming at the junctional zone with regular line features,the optimization of the junctional zone with regular line features is based on the rule line constraint formed by the semantic constraints of two regular walls.By the calculation of triangulation and space projection to triangles in the junctional zone,the triangle mesh in the junctional zone with regular line features is optimized.For junctional zone with regular surface features,the optimization of the junction area with regular surface features is based on the regular surfaces that are fitted according to the point cloud data of the junction zone.By the calculation of space projection and triangulation to triangles in the junctional zone,the triangle mesh in the junctional zone with regular surface features is optimized.Through comparing many experiment results of point cloud data of different buildings,the validity of the method in this paper is verified.Finally,the entire building model reconstruction process is integrated into the designed system to complete the whole process of optimizing the original point cloud of the building,and some auxiliary functions of model editing are added to make the whole building model reconstruction system more perfect.