【作者】 谢天；

【导师】 郑耀；

【作者基本信息】 浙江大学 ， 空天信息技术， 2013， 博士

【摘要】 增强现实技术是将计算机产生的虚拟信息与现实环境无缝融合,通过显示设备呈现给用户的一类视觉技术。它能在尽量不影响用户观察现实环境的条件下,给予用户诸如文字、动画等额外的信息,带给人们全新的视觉感受。鉴于其直观透明的交互方式和潜在的巨大发展空间,增强现实技术正受到世界各地学者们越来越多的关注,其应用领域也渐渐遍布娱乐、医疗、商务、教学、军事、工程制造等等各方各面。在这些应用中,有一类专门为操作者提供虚拟指引信息,辅助或指导用户实际操作的应用,本文统称其为面向操作指引的增强现实系统。这类系统包括维修诱导、工业培训、医疗仿真等等,随着增强现实技术的快速发展,正受到越来越多的关注。但目前大多数研究仍聚焦于注册、交互等关键技术的使用和突破上,对其通用的系统架构和信息组织方式却很少涉足。在这个背景下,本文从设计通用系统框架的角度出发,分析了面向操作指引的增强现实系统所涉及的各种技术难点。并针对这些难点,分别研究了基于视觉的注册技术,多通道交互的冲突管理,各技术模块的形式化建模,复杂信息的组织管理等等。提出了一整套面向操作指引的增强现实系统的通用架构和构建细节,并实现了两套原型系统用以验证。总体而言,本文的主要贡献如下：1)提出了一种基于平面颜色分布的自然标志三维注册配准算法。该自然标志注册算法提取场景中具代表性的彩色连通域,通过色相、倾角、面积、形状等特性对比匹配待跟踪模板,依据模板颜色分布的全局特性作优化筛选得出最优匹配组,最终计算出三维姿态。算法适用于颜色非单一的彩色场景,无关纹理是否复杂,且实时准确,适用于多种应用环境。2)提出了一种面向操作指引的增强现实系统的通用层次框架。该框架对增强现实或操作指引所需的各方面功能做出了合理的组织与分配,建立了该类系统的通用设计理论和通用模块规划,并对各功能模块都做了详细的功能描述。该通用框架能为今后该类系统的设计和实现提供思路和模板。3)针对本文提出的通用系统框架,使用高级Petri网对整个框架流程和各功能模块做了详细的形式化建模。明确了系统中各数据的流动关系,定义了每个流程细节具体的输入输出关系,也为系统的实际实现提供了软件设计层面的模型基础。本文还基于Petri网提出了一种处理交互冲突的优先选择策略模型,为该类系统中多通道交互的冲突管理提供了有效的解决方案。4)提出了面向操作指引的增强现实系统具体的信息组织方式。在S1000D技术出版物国际标准的基础上扩展提出了完整的系统通用数据格式和引用细则,并给出了各类数据模块的XML编著示例及索引方法。基于本文提出的这些数据规范和检索方法,注册信息和交互指令可以有机地融入工序信息和指引信息的数据结构之中,使工序和指引信息的检索能在明确的有向链接下快速实现,以保证系统的实时运行。更多还原

【Abstract】 Augmented Reality (AR) is a modern technique, which can merge virtual information into real world without any conflict when observed by a display device. It combines real and virtual objects within a single real environment, runs interactively in real time, provides auxiliary information such like texts or animations to users, and tries not to introduce any disturbance against the real environment. As AR technique offers intuitive interactions and relatively simple design requirements, it attracts more and more attentions from researchers all over the world. Now AR has been applied to many different areas including entertainment, medical attendance, business, education, military, manufacture, and so on. Among these applications there is one kind of systems such like maintenance guidance, industrial training, surgery simulation and so on, which are designed to guide operators in their work by showing various virtual guidance information. In this thesis, we defined this kind of systems as Augmented Reality systems catering to operation guidance. Since AR develops so fast in recent years, much more attentions have been made on this subject, and many similar systems have been developed. But most researches still concertrate on how to apply or improve AR’s enabling techniques such as registration and interation, while few has worked on general system frameworks or information organization.From the perspective of general system frameworks, this thesis analyzed potential technical challenges that are concerned in AR systems catering to operation guidance. Herein, techniques such as vision based registration, conflict management among multimodal interactions, formal modeling of each technical modules, complex information organization and retrieval, are studied int this thesis, and meanwhile a general framework with careful details of every particulars in AR systems catering to operation guidance is proposed as a paradigm. Two prototype systems were developed to demonstrate these techniques and principles.All in all, the main contributions of this thesis are listed as follows:·A nature feature registration algorithm based on color distribution is proposed. The algorithm takes colorful connected areas as invariant features, calculates their descriptors simply by hue and geometry information, and matches them by global optimization based on geometric constraints on color distribution under undefined view transformations. The algorithm doesn’t rely on complex features and textures. It is suitable for most environments except those who have uniform color almost everywhere. Experimental results demonstrate the efficiency and effectiveness of the proposed algorithm, which runs in real-time with good precision robustly.· A general system framework of AR systems catering to operation guidance system is proposed. The framework establishes a general design theory and a common functional division. Every specific functional module is arranged to satisfy a special function of the system, and each of them are designed in careful detail. Hope this general framework can provide basic principles of design and accomplishment of such systems.· High level Petri net is applied to build the formal model of the general framework including every functional module. The model confirms the workflow, defines the specific I/O relations, and offers basic software architecture. A Petri net-based strategy to handle the conflict caused by multimodal interations is also proposed, which provides effective solution to manage this unstable problem.· A novel data design layout is proposed to organize the complex data involved in AR systems catering to operation guidance. The layout is an extended design based on S1000D international specification for technical publications. It formulates four original types of data module, and establishes their own XML schemas with retrieval rules. Markup samples are given to test the ability of this layout, which demonstrates linking registration and interaction information directly to procedural and guidance data module is a efficient way to keep system running in real time.更多还原