Abstract: This paper presents a real-time markless tracking framework for augmented reality game, which can work in a texture-abundant environment without any prior knowledge of the scene. It can track a camera and several rigid bodies in the sight of the camera in a parallel way, attaining their pose and movement information. The research is based on a parallel tracking and mapping framework developed by the Oxford University, which can only work robustly in static environments. The results show that by utilizing the state-of-art multi-core processing technologies, tracking camera and moving objects in a dynamic environment is practicable. It presents a cheap, convenient and simultaneous solution to display and interaction issues, the two core problems involved in many AR applications.

Key words: Simultaneous Localization And Mapping（SLAM）, camera tracking, augmented reality, object detection

摘要： 针对增强现实游戏应用，介绍了一种基于自然特征的实时跟踪算法，它可以在一个未知但纹理丰富的环境中，不需要任何预先的学习，对相机和若干出现在相机视野里的刚体进行并行跟踪，获得相机和各个刚体的姿态和运动信息，可以在刚体移出相机视野以及相机剧烈运动导致跟踪丢失的情况下进行自动恢复。基于静态环境中并行相机跟踪和三维重建的算法框架进行扩展和改进，引入了用于交互的运动刚体。结果表明，利用现有的多核处理技术，在含有运动物体的环境中进行相机和运动物体的并行跟踪是可行的，为增强现实环境中同时支持显示和交互提出了廉价易行的解决方案。

关键词: 即时定位与地图构建, 相机跟踪, 增强现实, 物体检测