Research On Key Technologies Of 3D Seamless Reconstruction Of Multi-source Point Cloud Buildings Based On Space-ground Coordination

As a new method of surveying and mapping,three-dimensional laser scanning technology and oblique photogrammetry technology have been widely used in many fields,such as cultural relic protection,digital modeling,reverse engineering and so on.The laser scanning technique is favored by the practitioners because of its high accuracy and abundant information,and the oblique photography technique is widely used because of its high automation,convenient data acquisition and wide area.But in the process of collecting ground data and air data,there are inevitably some reasons,such as the limitation of view angle or the obstruction of obstacles,which result in the incompleteness of data collection and the cavitation and deformation on the surface of the model,thus,the difficulty of model construction is increased and the quality of model is decreased.Aiming at the situation that it is difficult to construct a complete and high-precision seamless model by single technology,the key technology of 3D seamless reconstruction of city is studied by using laser scanning and oblique photography to acquire the multi-source point cloud data of the object.The main work and achievements of the thesis are as follows:(1)The acquisition of laser point cloud data is introduced in detail,and some work is done to preprocess the laser point cloud data,and the verticality information is introduced to remove the special refraction noise;The method of obtaining image point cloud in the process of oblique image processing is also introduced.According to the difference of multi-source point cloud data,the characteristics of point cloud from different sources are analyzed.(2)The laser point cloud and image point cloud are registered.In order to improve the quality of cross-source point cloud registration,a new registration method based on PCA and Tr ICP is proposed.First,the source point cloud and the target point cloud are analyzed by PCA,and then transformed into the PCA coordinate system,finally,the cross-source point clouds are accurately registered by the Tr ICP algorithm which can adapt to different overlap rates.Experimental results show that this method is suitable for multi-source point cloud registration.(3)The method of feature extraction from laser point cloud and image point cloud is studied.Firstly,the normal vector estimation method of point cloud is introduced,then the edge contour feature of point cloud is extracted by Do N operator and vector angle difference,finally,the multi-facet feature of building is extracted by improved RANSAC algorithm,then,the edge contour is extracted from the building surface,and then the linear feature is extracted by using RANSAC algorithm to fit and project.(4)The method of building model based on multi-source point cloud data is studied,and then Sketchup is used to build parametric model of multi-source point cloud.