融合Kinect点云与光学影像数据的室内三维建模

发布时间：2018-01-29 00:09

本文关键词： Kinect Agisoft Photoscan软件数据融合三维建模　出处：《中国地质大学(北京)》2017年硕士论文　论文类型：学位论文

【摘要】：Kinect是微软推出的融合彩色相机与深度相机的传感器,最初应用于体感游戏中。Kinect在获取彩色影像的同时也能获取三维激光点云数据,相比传统的三维激光扫描仪,具有价格低廉,操作简便等特点,因此被研究者广泛的应用到了小场景的三维建模工作中。但与此同时,Kinect也具有其不可避免的缺点,其采集数据时深度感应范围有限,并且反射值高的部位无法得到点云,对于点云的拼接工作量大,传统的点云拼接算法无法有效的应用到Kinect获取的点云数据中。本文是在融合了Kinect点云与光学影像数据的基础上进行的室内三维场景的建模。首先利用Agisoft Photoscan软件对Kinect获得的彩色光学影像进行处理从而获得大量三维点,将其与拼接完成后的Kinect点云进行融合,经过融合后的点云质量有明显改善,弥补了Kinect点云缺失严重的现象。然后分别针对室内场景中不同的对象利用不同的方法对融合后的点云数据进行分割等处理,并通过“孔洞算法”提取其边界点,最终将边界点进行规则化拟合处理从而完成室内三维场景模型。本文的主要工作和创新点包括:1.搭建Kinect for windows2.0平台,采集室内场景的彩色影像及点云数据,并且通过Agisoft Photoscan软件将彩色影像转换成点云数据2.提出了融合Kinect和Photoscan点云数据的方法,提高点云精度,为室内三维场景建模提供数据支持3.分别采用RANSAC和区域面生长法对融合后的点云数据进行分割,分析比较了两种算法的优缺点4.研究了基于点云对室内三维场景中不同物体通过不同方法进行几何建模的方法,并对线框模型进行纹理建模。实验结果表明,本文所提出的融合Kinect点云和光学影像数据进行室内建模的方法,能够提供质量较高的三维点云数据并且能得到较好的室内三维模型,与传统设备相比不仅降低了三维建模的费用还提高了建模的自动化程度。
[Abstract]:Kinect, a sensor from Microsoft that combines color cameras with depth cameras, was originally used in somatosensory games to capture color images as well as 3D laser point cloud data. Compared with the traditional 3D laser scanner, it has the characteristics of low cost and easy operation, so it has been widely used in the 3D modeling of small scenes, but at the same time. Kinect also has its inevitable shortcomings, its depth sensing range is limited when collecting data, and the point cloud can not be obtained in the part with high reflection value. The traditional point cloud mosaic algorithm can not be effectively applied to the point cloud data obtained by Kinect. This paper is based on the fusion of Kinect point cloud and optical image data to model the indoor 3D scene. First use Agisoft. The color optical image obtained by Kinect is processed by Photoscan software to obtain a large number of 3D points. After fusion with the Kinect point cloud after splicing, the quality of the point cloud is improved obviously. It makes up for the serious lack of point cloud in Kinect. Then it uses different methods to segment the fused point cloud data for different objects in the indoor scene. The boundary points are extracted by "hole algorithm". Finally, the boundary points are regularized to complete the indoor 3D scene model. The main work and innovation of this paper include: 1. Build the Kinect for windows2.0 platform. Collect indoor scene color image and point cloud data. And the color image is transformed into point cloud data by Agisoft Photoscan software. 2. The method of merging Kinect and Photoscan point cloud data is proposed. Improve the accuracy of point cloud, provide data support for indoor 3D scene modeling. RANSAC and area surface growth method are used to segment the fused point cloud data. 3. The advantages and disadvantages of the two algorithms are analyzed and compared. 4. The geometric modeling method of different objects in indoor 3D scene based on point cloud is studied, and the texture modeling of the wireframe model is carried out. The experimental results show that. The method of integrating Kinect point cloud and optical image data for indoor modeling can provide high quality 3D point cloud data and obtain better indoor 3D model. Compared with traditional equipment, it not only reduces the cost of 3D modeling, but also improves the automation of modeling.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212;TP391.41

【相似文献】