3D Reconstruction Of Buildings Based On Lidar Point Cloud Data

With the improvement of 3D data acquisition ability,people’s demand for 3D data reconstruction is increasing day by day.3D reconstruction of buildings is the main content of digital city research which including the reconstruction of 3D shape and surface texture of buildings.The 3D reconstruction of buildings is of great significance in enhancing the acquisition ability of building digital models,the restoration ability of ancient buildings and the establishment of military digital battlefields.At present,there are three key problems in the research of building reconstruction.First,point cloud information contains all kinds of target point cloud information and noise point cloud information.It is necessary to effectively classify all kinds of point cloud information and remove noise data.Secondly,most of the reconstruction of the building surface focuses on the reconstruction of the facade topology,without an effective expression of the building surface details.Thirdly,the registration of point cloud data and image data is still the key to the fusion of the two data sources.In practical applications,due to the different attributes and acquisition methods of point cloud data and image data,there exists deviation between point cloud data and image data on time,which makes the fusion of the two data sources very difficult.In view of the above three key issues,the following studies were carried out:(1)The multi-body dynamic error analysis model of the vehicle-mounted LiDAR scanning system was established to analyze the error of the vehicle-mounted LiDAR scanning system,and the parameters of the vehicle-mounted LiDAR scanning platform were determined according to the error range.In addition,the GPS-based point cloud splicing technology unifies the point cloud data of many different measuring positions into the geodetic coordinate system to complete the point cloud data splicing.(2)A noise-eliminating algorithm based on the convexes of the α van number constraints converge is put forward.After optimizing average filtering algorithm,this essay filtering the normal vectors of original point cloud data.These filtered normal vectors can represent the coordinated local surfaces,regard these surfaces as convex set,and to utilize algorithm on α van number constraints converge to upgrade the positions of points in point clouds.Meanwhile,this essay also presents a segmentation model,which is based on the potential energy function of point clouds,uses bandwidth parameters which are produced during the potential energy function adjusts point cloud clusters,and change bandwidth parameters as well as divide point cloud data in the process of convergence of potential function.(3)The paper proposed a building facade area level depth reconstruction method.The method is based on the building facade point cloud to arrange the point cloud by building facade principal plane model.We can also use r-edge method to abstract and rule the edge of point cloud plots and to utilize area layer ways to classify the information on facades of architectures.By establishing the mathematical model building facade area block depth,the depth problem for the label assignment math problems.Therefore,effectively block depth determines the building facade area,to ensure the integrity of the building facade details and topological relations.(4)A method about registration between point cloud data based on edge characteristics and on drone video is put forward.We can build a laser point cloud 3D model and drones in a coordinate system as well as their relationships,in order to gain the same name edge information and build mathematics model on the same name edge information.The paper also established the mathematical model of the corresponding edge information matching the model is solved by completing 3D point cloud model with drone video image registration work.Through experimental analysis on the problems mentioned above,both noise-eliminating algorithm based on the convexes of the α van number constraints converge and the potential energy function of point clouds have an obvious effect on eliminating noise point,clearly classifying and other characteristics.The reconstruction method based on the hierarchical depth model of the building facade can not only reconstruct the topological structure of the building facade,but also reconstruct the details of the building facade.By analyzing the precision of building surface texture,the validity of point cloud data and drone video image registration method based on edge feature is verified.The comprehensive experiments show that the method can be used to reconstruct the 3D model of complex buildings quickly.