Research On Segmentation Of Scattered Point Cloud And Identification The Feature Surface

Three-dimensional laser scanning technology has become an importanttechnology in modern surveying and mapping based on its fast, high precision andother characteristics. Nowadays, Three-dimensional laser scanning technology hasreceived a large number of applications in the digital of ancient buildings,deformation observation of large buildings and reverse engineering. Compared tothree-dimensional laser scanner quickly receive vast amounts of high-precisionpoint cloud data, the associated data processing method is still not perfect. Such asa common ancient building that contains plates, beams, columns, brackets andother components maybe contains tens of millions points. So, how to storage,management, transmission, modeling and analysis applications the data quickly is aserious problem we faced. How to simplify complex issues, and divided complexcomponents into several simple components. In this context, this paper discussesthe segmentation process of point cloud and identification of some point cloud thatfrom rules characteristics surface.This paper analyzes the strengths and weaknesses of commonly approachesused to building index of the point cloud, and then combined advantages of gridand octree to building a new index: grid-octree. We got the vector through fittingpoint cloud into planar, and improved the method of adjust normal vector that onceoften fall into deadlock problems, and improve the efficiency of the law to adjust.We researched some frequently used for a variety of differential geometry of thecalculation method, and the corresponding algorithm to achieve. Focus on the mostcommonly used variety of point cloud data segmentation in-depth discussions,analysis of the scope of application of various methods and for different types ofpoint cloud segmentation choose a different division. The common variety of thecurrent edge point cloud point extraction methods were studied, analyzed variousdata grid algorithm for extracting the edge effect, then the introduction theprojection method of the nearest point of the grid center to determine to solve theedge of the grid identification error, then accurate to extract edge points fromcomplex object point cloud model. Finally, we analysis the geometriccharacteristics of the point cloud that come from plane,spherical and cylindrical, and achieve surface feature identification and parameter calculation of the threerules point cloud.According to the related researches, we designed and developed system forexperiment using object-oriented language C#and OpenGL3D graphics toolkit inVisual Studio2010software development platform. For the scattered point clouddata that only contain X, Y, Z three-dimensional coordinate information, the use ofprogramming methods to achieve the point cloud data in3D display spaceperspective transform, scaling, rotation and translation and other basic operatingfunctions. With three-dimensional observation operation and display inexperimental system, so the index of point cloud data, normal vector of point cloud,curvature of point cloud and the results of point cloud segmentation were observedand verified. This demonstrates the feasibility of the algorithm. And the finalresults of the data to be saved.