【作者】 姜晨；

【导师】 卢荣胜；

【作者基本信息】 合肥工业大学 ， 测试计量技术及仪器， 2005， 硕士

【摘要】 基于机器视觉理论的三维视觉检测技术已经成为在线检测领域中的一个重要的测量技术。通过图像重构出被测物体的三维坐标是三维视觉检测中必不可少的过程。在三维视觉场景重构过程中,一个众所周知的难题就是场景图像上的坐标点匹配问题。通常解决该问题的一种有效的方法是采用结构光主动视觉技术,如点结构光、线结构光扫描法以及编码结构光法等。要想用一幅图像在真正的三维欧氏空间重构三维场景,尤其是动态的三维场景,最有效的方法是采用编码结构光照明主动视觉技术及其装置。因此,研究合适的编码结构光照明方法,解决场景图像上的彩色伪随机编码模板的坐标点匹配问题,是三维场景中的一个极其重要的理论与技术问题。 本论文主要研究伪随机彩色空间编码三维场景特征化技术。即通过对三维场景表面进行伪随机彩色空间编码,并根据伪随机编码的一个重要属性——窗口特性,使三维场景表面上的每一个采样点被唯一辨识,同时通过Harris角探测提取采样点的坐标,解决三维表面重构时图像解码难题。归纳起来,论文主要的内容和成果如下: 一 介绍伪随机彩色编码阵列的原理,总结伪随机彩色编码照明方法与其他经典照明方法的差异,同时编制相应的伪随机阵列。 二 结合特征点提取算法,讨论彩色伪随机编码模板的相关特性,包括编码颜色、形状、密度。 三 实现基于Harris角探测和模板匹配法的彩色伪随机编码模板的图像特征点提取算法。更多还原

【Abstract】 3D (three-dimensional) vision inspection based on machine vision theory has become an important measurement technique in on-line measurement field. The essential procedure of 3D vision inspection is to reconstruct the 3D dimensional coordinates of the object to be inspected from its images. In this process, however, there is a known difficult problem, i.e. the registration of correspondences between object surface and its images, and among images, especially for the free-form surface 3D reconstruction. In order to tackle this problem, active 3D vision inspections with structured light illumination, such as dot structured light, line structured light and encoded structured light pattern, are often employed in practical industry measurement. Among the varieties of structured light illumination, the most effective one is the encoded structured light pattern, because the 3D vision inspection system with this illumination method not only can be used for reconstructing the 3D coordinates of free-form surfaces, but also for dynamic or variable 3D scenes. Therefore, the research of the encoded structured light pattern to resolve the correspondence problem is very important in 3D vision inspection.In this paper, we concentrate our mind on a promising pseudo-random color encoded illumination pattern and its application. The illumination pattern has one of important characteristics, window property. Any feature point of the illumination pattern after emitted on the surface can be identified exclusively with respect to the window property of the pseudo-random color encoded array. The coordinates of the feature points on images can be obtained by Harris corner detector. Thus the correspondence problem depicted above can be readily solved.In the paper, the principle of the pseudo-random color encoded array, its illumination pattern feature extraction and research achievements in question are detailed. In summary, they are listed as follows:1. The principle of pseudo-random color encoded array is introduced. The differences of the illumination pattern from other illumination patterns are summarized. The program of the array pattern is given.2. The characteristics of the pseudo-random color encoded pattern with respect to its elements: color, shape and density of code are discussed.3. The algorithms of the image coordinate extraction of the pseudo-random color encoded pattern based on Harris corner detect and model matching method are implemented.更多还原