Reconstruction Of 3D Terrain Based On Large Scale Point Cloud Data And Its Parallelization

The surface reconstruction of three-dimensional terrain is a technique to visualize the real surface morphology as a three-dimensional visual model in the computer;the current popular method is to reconstruct the surface model based on the point cloud;the main means to obtain the terrain point cloud is in addition to the traditional In addition to Lidar scans,there are also feature extraction methods based on oblique photogrammetry measurements.The latter has gained more and more attention due to its advantages of cheaper operation methods and more flexible processing algorithms;at such points On the basis of cloud data,combining such demand conditions,it is particularly important to choose a high-quality reconstruction method that is compatible with the terrain point cloud.This paper starts with the surface reconstruction of large-scale point cloud,analyzes two reconstruction methods based on Delaunay triangulation and Poisson formula,compares their performance on terrain reconstruction,analyzes their advantages and disadvantages,and finally selects Poisson.The reconstruction algorithm carries out the key research and analysis.Based on the existing reconstruction algorithms,the parallelization calculation is added to adapt to the data characteristics of the large-scale terrain point cloud,in order to improve the performance of the algorithm on this type of data.The main research tasks in this paper are:(1)Combing the concept of topographic point cloud,classifying the method of obtaining topographic point cloud,and providing a representative point cloud processing library application interface according to the applied format;(2)From the mathematical aspect The Delaunay triangulation and Poisson reconstruction methods are analyzed from two aspects of the algorithm and the level of the algorithm.The advantages and disadvantages of the two reconstruction methods are theoretically studied and their applicability to terrain point cloud data reconstruction is discussed.At the same time,two algorithms are actually compared.(3)On the basis of studying and understanding the principle of parallelism,choose to improve the parallelization of Poisson reconstruction,and complete the Poisson algorithm parallel scheme according to the characteristics of terrain data;(4)Build the corresponding platform Select 700,000 class point clouds as experimental data to view the actual situation of the parallelization algorithm,which mainly includes whether the algorithm after parallelization is lost in the details of the model and whether the time cost is reduced.