【作者】 牟冰；

【导师】 葛宝臻；

【作者基本信息】 天津大学 ， 光学工程， 2004， 硕士

【摘要】 随着社会的进步与科学的发展，特别是计算机技术和数字化技术的飞速发展，人们越来越多的关注三维物体的数字化技术。三维数字化技术可以广泛应用于机械制造与设计，人体工学测量、服装设计，医学，文物保护等领域。对于三维数字化系统来说，现实世界三维信息到计算机虚拟世界的信息转换的映射关系，是通过标定工作实现的，标定技术是三维数字化技术中最为关键的技术。针对不同的数字化系统，围绕各种标定方法的研究是目前的一个热点，而且，为实现物体的真彩色模型可视化，彩色信息与空间三维信息匹配也是标定技术的研究内容。本文围绕这些问题，利用BP神经网络算法完成了三维彩色数字化系统的标定，初步实现了空间信息和彩色信息的匹配，主要工作及创新性如下：研究了激光三维传感器和激光彩色传感器的数学模型和标定理论，提出了利用细丝靶标的BP神经网络标定方法。设计并建立了三维数字化实验系统，根据提出的标定方法采用四层BP神经网络进行了标定实验研究，得到了BP神经网络的结构参数，完成了传感器的标定。对实际彩色物体进行扫描实验，对标定理论和具体方法进行了验证，再现得到了三维彩色点云图，初步实现了空间信息和彩色信息的匹配。更多还原

【Abstract】 Along with the society advancement and science development, especially the rapid development in computer technique and digitization technique, technique related to the digitization of 3D object catches more and more attention. 3D digitization technique can be widely applied in the fields of mechanism manufacture and design, body measurement and dress design, medicine, cultural relic protection and so on. In 3D digitization system, the mapping relationship between real world 3D information and computer virtual world information is derived by calibration. Calibration technique is the most important technique of 3D digitization technique. Different calibration methods towards different digitization systems are highlights at present. Furthermore, in order to realize the visualization of 3D color object, the matching of color information and space 3D information is also the study content of calibration technique.Concentrating on these problems, the calibration of 3D color digitization system is done by BP neural network arithmetic, and the basic matching between space information and color information is realized. The main study and innovation are as follows:The mathematics model and calibration theory of laser 3D sensor and laser color sensor has been studied. Calibration technique based on BP neural network was presented, which use a filament target as calibration target.3D digitization experimental system was designed and built. Calibration study was carried out by using a four-layers BP neural network and structure parameters was got from the calibration.Real color object was scanned to verify the calibration theory and methods. 3D color point cloud was obtained, and the basic matching of space information and color information is realized.更多还原