【作者】 贺宇；

【导师】 徐晓东；

【作者基本信息】 内蒙古大学 ， 计算机技术， 2019， 硕士

【摘要】 目前计算机的硬件和软件都得到了飞跃式的发展,三维重建技术相比过去也得到长足的进步。三维重建的目的是为物体或场景建立可用于计算机处理和分析的三维模型。目前,由扫描仪设备或者基于图像的三维重建方法获得的原始数据均为三维点云数据,但是在3D打印、虚拟现实、数字城市、智慧交通等相关应用中均需要目标对象的多面体表面模型或者CAD实体模型。由三维点云构建三维模型是三维重建的关键技术之一,也是逆向工程的核心问题。通常,狭义的三维重建是指由三维点云构建三维实体模型的过程。在实际应用中,由于通过三维扫描、图像处理等方法采集的三维点云是包含多个物体的复杂数据,因此以物体为单位对点云进行分割处理是三维重建过程中必要的预处理手段。目前点云分割处理主要依靠相关专业软件人工完成,存在处理成本高、效率低等问题。本论文主要研究复杂点云数据的自动分割方法以及在三维重建中的应用。本文提出一种基于平面模型的分割方法,该方法利用逐步分割的思想。首先在三维点云中拟合平面模型,统计三维点云在平面模型的位置情况,保留下只在平面一侧有三维点云的平面模型作为点云的边界平面模型,将边界平面模型从初始点云中分离,从而产生有明显空隙的三维点云。然后利用这些空隙对三维点云分类。最后在每一类点云中拟合平面模型并与边界平面模型平行,作为点云的切割面,实现在物体堆叠条件下的分割方法。本文通过实验证明本方法存在一定的有效性。更多还原

【Abstract】 At present,the hardware and software of the computer have been developed by leaps and bounds,and the three-dimensional reconstruction technology has made considerable progress compared with the past.The purpose of 3D reconstruction is to create a 3D model for computer processing and analysis for objects or scenes.Currently,the raw data obtained by the scanner device or the image-based three-dimensional reconstruction method is three-dimensional point cloud data.However,in the related applications of 3D printing,virtual reality,digital city,intelligent transportation,etc.,the polyhedral surface model or CAD solid model of the target object is required.Constructing a 3D model from a 3D point cloud is one of the key technologies for 3D reconstruction and the core issue of reverse engineering.In general,narrow threedimensional reconstruction refers to the process of constructing a three-dimensional solid model from a three-dimensional point cloud.In practical applications,since a three-dimensional point cloud collected by a method such as three-dimensional scanning and image processing is complex data containing a plurality of objects,Therefore,segmentation of point clouds in units of objects is a necessary preprocessing method in the process of three-dimensional reconstruction.At present,point cloud segmentation processing mainly relies on relevant professional software to complete manually,and there are problems such as high processing cost and low efficiency.This thesis mainly studies the automatic segmentation method of complex point cloud data and its application in 3D reconstruction.This paper proposes a segmentation method based on planar model,which uses the idea of stepwise segmentation.Firstly,fit the plane model in the 3D point cloud,and count the position of the 3D point cloud in the plane model.The plane model with a three-dimensional point cloud on one side of the plane is reserved as the boundary plane model of the point cloud,and the boundary plane model is separated from the initial point cloud,thereby generating a three-dimensional point cloud with obvious gaps.Then use these gaps to classify 3D point clouds.Finally,the plane model is fitted in each type of point cloud and paralleled with the boundary plane model.As the cutting plane of the point cloud,the segmentation method under the object stacking condition is realized.This paper proves that the method has certain effectiveness through experiments.更多还原