移动增强现实大范围定位与追踪关键技术研究

Equipped with high-resolution camera and color displays, mobile phones have evolved into a powerful platform for Augmented Reality (AR) applications. Wide area location recognition and camera tracking are the two key problems currently limit the usability of Mobile Augmented Reality (Mobile AR) systems. Compared to the desktop computers, the phones' small size and lightweight make them extremely portable. Thus, the requirements of wide area location recognition and camera tracking are more urgent for mobile AR than for traditional PC based AR applications. .However, the above problems are far from being solved because neither computing power nor memory space of the resource constrained mobile phones can well satisfy the requirements of the wide area on-device mobile AR applications. In view of this, we will make following contributions to solve the above difficulties. .Firstly, we propose to partition the whole scene into several sub-maps according to the user's preference or the requirements of the mobile AR applications. Since the built sub maps are geometrically independent to each other, we need only to load the map currently being tracked instead of the global map into the memory to realize wide area mobile AR. So, the multiple maps method is more flexible in memory management and especially suitable for resource constrained mobile AR applications..Secondly, we propose an efficient vector encoding strategy and design a fast approximate nearest neighbor search method to implement a flexible image search engine for the use of on-device location recognition. While compact enough for mobile AR applications, the method can also reduce the computational time obviously when combining with our visual and GPS integrated indexing technique. .Thirdly, we also propose a flexible natural features describing and matching method to get a real time camera initialization and tracking system. While fast and compact enough for mobile AR, the method can also provide more accurate matching results, which facilitates the real time augmentation process on low power mobile phones to a large degree. .We will build a prototype system and apply it to real projects to prove the effectiveness of the proposed methods. We believe that the implementation of this research has important theoretical and practical significance for the development of mobile AR techniques.

大范围场景下的定位与追踪是移动增强现实（简称移动AR）所面临的一项亟待解决的关键问题。然而,当前诸如智能手机之类的移动计算设备普遍存在资源受限问题，并不能很好的满足移动AR大范围定位与追踪对运算能力以及存储空间的需求。对此，本课题将通过深入分析基于子场景的大范围场景表达对移动AR定位与追踪性能改变的机制，构建解决资源受限模式下的移动AR大范围定位与追踪问题的理论和方法体系。在此基础之上，拟研究图像视觉描述与量化编码、视觉与地理位置信息相结合的搜索区域划分与索引构建、基于快速近似最近邻搜索的海量图像检索以及移动设备摄像机位姿快速初始化与追踪等相关问题的解决方案。本课题还将通过构建原型系统并结合具体的项目来验证所提出的理论框架与解决方案的正确性和有效性。本项目的实施对于移动AR的实现、发展以及广泛应用具有重要的理论和现实意义。

针对移动AR中的大范围定位注册问题，本课题通过深入分析基于子场景的大范围场景表达对移动AR 定位与追踪性能改变的机制，构建了解决资源受限模式下的移动AR 大范围定位与追踪问题的理论和方法体系。在此基础之上，研究了图像视觉描述与量化编码、视觉与地理位置信息相结合的搜索区域划分与索引构建、基于快速近似最近邻搜索的海量图像检索以及移动设备摄像机位姿快速初始化与追踪等相关问题。经过四年的研究工作，课题组圆满完成了既定目标，共发表论文11篇，其中4篇发表在IEEE、ACM等国际重要刊物上。

内容获取失败，请点击重试