【作者】 易凯；

【导师】 姜光；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2010， 硕士

【摘要】 三维测量及配准技术是计算机视觉技术的一个分支,是计算机视觉和计算机图像图形处理相结合的一个研究方向,它在生产自动化、机器人视觉、文化遗产保护、虚拟现实和计算机辅助医学等领域都有着广泛的应用前景。在三维测量及配准技术中,如何寻找一种即简便又快捷的实现方式是一项重要而有难度的研究课题。本文在深入研究结构光三维测量相关理论、方法及关键技术的基础上,设计与开发了基于格雷码结合相移技术的双目立体结构光三维测量系统。在分析总结了目前国内外通常所采用的点云数据配准方法的基础上,形成了本文中所采用的配准流程。最后,在实验室环境下利用自行搭建的硬件平台与开发的软件图像处理系统,进行了可行性方面的实验。首先,本文阐述了双摄像机结构光法实现的物体三维重构的原理,然后建立了该方法的数学模型。其次,重点论述了系统的构成和双摄像机定标,得到了摄像机的内外参数和编码图案。再次,本文论述了三维点云重构的软件设计,根据图像处理算法,对编码图案进行图像信息的提取、滤波和二值化,利用编码图案、外极线约束和三角测量原理,获得物体三维坐标。最后,在研究经典ICP算法的基础上,利用点到点的ICP算法实现点云数据配准,得到了较好的收敛效果。更多还原

【Abstract】 Three-Demension measurement and registration is a branch of computer vision. It is a research direction that integrates computer vision with computer image and graph processing.It has a wide application in the fields of automation process, machine vision, culture relic conservation, virtual reality, computer aided medicine and so on. In 3D measurement and registration technology, a fast and easy realization way is a big problem.Based on deep study of relevant theories and key technologies in the dissertation, a 3D measurement system based on combining Gray Code with Phase Shifting method using double CCD structured light is developed. After analyzing and comparing the widely used point cloud data registration methods, we came to a registration pipeline. At last,hardware platform is built and software of image processing system is developed in lab environment, and we experimentalize the idea.Firstly, the principle of double CCD strctured light 3D measurement methed is presented. Then the mathematical model using this method is established. Secondly, the structure of the system and the dual-camera calibration are introduced. The parameters of the camera and the coded pattern pictures are obtained. Thirdly,the software flow is established. The whole image processing flow chart include stripes extracting,image filtering,binary transformation.Then the 3D coordinate can be obtained using coded pattern pictures, epipolar constraint and the principle of optical triangulation. Finally, based on the study of the classical ICP algorithm, the registration of point clouds is implemented using point-to-point ICP algorithm. A good convergence is gained.更多还原