Feature Extraction And Modeling Of Building Point Cloud Based On Morse Theory

Light Detection and Ranging(Li DAR)is a new and efficient 3d space acquisition technology which has developed rapidly in the past two decades.It can directly obtain 3d space information on the surface of ground objects.The data structure obtained by Li DAR is in the form of object surface point cloud,which consists of many dense or sparse single points,including 3d coordinates,reflection intensity and color information of objects.It is the basic data source for reverse engineering(Reverse Engineering)model reconstruction,and provides a new technical means for 3d digital city research,especially building model reconstruction.As the most important artificial ground object in the city,the building has the characteristics of large number,large density,complex spatial structure and wide variety of forms,so it is of great significance to carry out 3D model reconstruction.But the high density of point cloud data also brings the burden to the computer processing,only hopes to increase the computer operation and processing speed,increases the data storage space already difficult to meet the actual engineering application demand.Therefore,the study of how to extract key feature points from building dense point cloud and realize the simplification and expression of topological feature information on building surface can not only reduce the amount of basic data,but also provide concise and effective feature data for building model reconstruction,but also improve the overall speed of computer operation and model reconstruction.The Morse theory used to express the topological information of two-dimensional scalar field is applied to the surface feature extraction of three-dimensional building model.The key problems,such as feature point extraction and classification,Morse-Smale dual complex,single complex and simplification,are studied and discussed.The main research contents include the following aspects:1)Combing and summarizing the basic theory,existing research results and existing problems.Described the research background and significance of this paper,summarized and summarized the building feature extraction and model reconstruction,point cloud model feature extraction and simplification,Morse theory in three-dimensional feature extraction domestic and foreign research status,the theoretical basis of this paper classic "Morse theory" is introduced emphatically.2)Building point cloud feature information extraction based on Morse Theory.Aiming at the low degree of automation in the traditional building point cloud feature extraction process and the lack of topological relationship of feature lines,a point cloud feature extraction algorithm based on Morse theory is proposed.Firstly,by analyzing the topological relation of the triangular network constructed by point cloud,the algorithm obtains the topological adjacent relation between the midpoint and the point,the point and the triangle,and the normal vector variation of the vertex in the triangular mesh is defined as the theoretical function index of piecewise linear Morse.The feature points are divided into saddle point,maximum point and minimum point.Different kinds of feature points are connected to construct different feature lines.The saddle point is connected to the maximum point to construct descending Morse complex,and the saddle point is connected to the minimum point to construct ascending Morse complex,ascending complex superimposed with the descending Morse to form Morse-Smale dual complex.Based on this algorithm,the feature extraction experiments of two groups of building surface model data are carried out.Firstly,all kinds of feature points in the model are obtained,and the corresponding Morse-Smale dual complex is constructed to realize the complete segmentation of the model surface.3)An algorithm for constructing single complex and simplifying.By applying the piecewise linear Morse theory from the two-dimensional surface model to the three-dimensional building surface model,the minima distributed in the original two-dimensional model are mostly distributed in the plane region of the non-featured information in the three-dimensional model.As a result,only the saddle points on the surface of the model are connected with the maximum points to construct feature lines,and the concept of single complex topology model is put forward,and the original simplified algorithm based on Morse-Smale dual complex is no longer applicable,andthe single complex topology model is put forward based on the original simplified method.The larger the simplified threshold,the fewer feature lines retained by the model,the more concise the model,the smaller the simplified threshold,the more features retained and the finer the model.4)Design and development of "Morse theory-based building point cloud feature extraction and simplified experimental system".Using C++ language,the proposed algorithms and key technologies are integrated into the prototype experimental system Microsoft Windows environment.The main functions of the experimental system include the import of raw data,the analysis of topological adjacent relations,the extraction and classification of feature points,the construction of Morse-Smale dual complex,the construction and simplification of single complex,and the statistics of the number of features.At the same time,the feature extraction and simplification experiment of the point cloud data of the Roman column model of the building and its accessories are carried out.The results show that the model feature lines are compact,clear and continuous.It is shown that this algorithm can not only be used for building models with obvious regular contour feature information,but also obtain relatively good results for 3D point cloud models with transition features such as arc curves.