Research On 3D Point Cloud Data Processing And Visualization Based On PC Cluster

In recent years,science and technology continue to advance,,two-dimensional urban information has been unable to meet people's needs while 3D city construction came into being.As an effective method for modeling,3D laser scan technology can quickly obtain high precision point cloud data,which has been widely used in the field of the protection of cultural relics,urban planning,reverse engineering and so on.In order to get the complete information of the object,we need to scan the data of different viewpoints to registration.The registration of point cloud data is the key process of 3D laser scanning data processing.However,with the increasing amount of data,the processing method of single computer can't meet the requirements of data processing speed,and the parallel computing method can deal with large data quickly and efficiently.Therefore,using the parallel method is an effective solution to deal with the 3D laser scanning data.Based on the above background,take the registration of 3D point cloud data as the research point,and improved the traditional ICP algorithm by using the geometric characteristics of the data.Furthermore,the parallelism of the registration method is analyzed.According to the parallel design method,the registration of multi view data for the appropriate registration scheme is carried out.Finally,the fast visualization of 3D model is discussed.The main research work is as follows:1)The multi view 3D data registration method was studied and optimized,and constructing the initial matching points by using the geometric features of the point cloud.Not only the matching information of the corresponding points,but also the characteristic information of the corresponding points are considered.Then the point pair is simplified to a certain extent,reducing the amount of data,and the error between the point clouds is eliminated quickly.Then using the ICP algorithm for fine registration.Experiments show that this method can improve the efficiency and accuracy of the registration process.2)Using the high performance parallel computing of PC cluster,combined with the MPI registration method with parallel program,and parallel decomposition of data and programs.Using a similar ring registration scheme for multiple view data registration,to design suitable for registration method for multi view point cloud data,in order to deal with the more rapid and efficient.3)To solve the single 3D model display load slowly,can built the level of detail model of the 3D model.According to the variati on of point of view,combined with distance selection method and eccentricity selection method to divide the screen,hierarchical loading on the 3D model in accordance with the classification scheme,to rapid visualization of 3D model.