CASIA OpenIR  > 毕业生  > 博士学位论文
面向增强现实的人体运动跟踪定位系统技术
其他题名Human Motion Tracking Technologies for Augmented Reality
杨浩
学位类型工学博士
导师吴福朝 ; 叶军涛
2011-05-28
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
关键词增强现实 人体运动跟踪 相机-惯性测量单元相对位姿标定 扩展卡尔曼滤波器 无迹卡尔曼滤波器 Augmented Reality Human Motion Tracking Camera-imu Relative Pose Calibration Extended Kalman Filter Unscented Kalman Filter
摘要人体头部和手臂位姿跟踪是增强现实系统的关键技术之一,头部位姿跟踪是增强现实中虚实融合绘制的基础,手臂跟踪对于实现良好的人机交互体验有着至关重要的作用。本论文的研究目标是实现大范围(约100m^2)室内环境中人体头部和手臂实时运动跟踪,为增强现实系统中的虚实融合绘制和人机交互提供必要的技术支撑。针对上述需求,本文首先研究了相机-惯性测量单元相对位姿标定,进而对室内环境中六自由度位姿计算、人体上肢运动实时跟踪等方面进行了比较系统和深入的研究,同时完成了一个原型系统。主要研究内容包括: (1) 提出了一种综合利用无迹卡尔曼滤波器和扩展卡尔曼滤波器的相机-惯性测量单元相对位姿标定方法。该标定方法简单易用,仅需要一个标定板即可完成,并且具有很好的标定精度和鲁棒性。 (2) 利用相机和惯性测量单元,设计了一个基于近红外LED标识点的六自由度位姿跟踪系统VisTracker-IR。通过三维重建和增量迭代优化的方法对标识点的三维世界坐标进行计算,利用(1)中的方法对相机-惯性测量单元的相对位姿进行标定。然后利用扩展卡尔曼滤波器对相机和惯性测量单元的测量数据进行融合,实现了六自由度位姿的实时跟踪。 (3) 利用多个惯性测量单元,设计了一个人体手臂运动跟踪系统LimbIMUTracker。提出了一种基于扩展卡尔曼滤波器的惯性测量单元-骨骼相对姿态标定方法。利用标定的惯性测量单元-骨骼相对姿态矩阵、惯性测量单元的输出姿态矩阵以及手工测量的骨骼长度,能实时估算手部在肢体局部坐标系中的位姿。 (4) 设计了一个面向增强现实的人体头部和手臂运动实时跟踪系统ARTracker。ARTracker系统包括:用于虚实融合绘制的人体头部定位子系统VisTracker-barcode和用于人机交互的人体手臂定位子系统LimbIMUTracker。ARTracker系统利用一个头部惯性测量单元对VisTracker-barcode和LimbIMUTracker的输出结果进行融合,实时计算手臂和头部在室内世界坐标系中的位姿。
其他摘要Motion sensing and tracking of the human head and arm/hand are of great importance for Augmented Reality (AR) systems. Six-DOF head tracking is essential for rendering augmented reality scenes, while six-DOF arm/hand tracking plays an important role for human to interact with the virtual environment. The research objective of this dissertation is to implement a human motion tracking system for large indoor environment, so as to provide a solid technical support for rendering a virtual object into a real scene, as well as human-computer interaction in the AR systems. To meet the requirements of certain AR applications, we designed a system to track the human head motion and arm/hand motion. The head motion tracking is based on a combination of a camera and an Inertial Measurement Unit (IMU), while the arm/hand tracking is based on multiple IMUs. The dissertation focuses on the relative pose calibration for Camera-IMU, relative pose calibration for IMUs and human body, as well as the fusion of data from multiple resources. The main contributions of this dissertation are: (1) A robust method, based on Unscented Kalman Filter (UKF) and Extended Kalman Filter (EKF), is presented for determining the relative pose between a camera and an IMU. The proposed method needs nothing but a simple calibration board. The simulation and experimental results show that the proposed method is effective and precise even if dramatic initial systematic errors and high noises are presented. (2) For an indoor environment decorated with infrared LED markers, a six-DOF motion tracking system, VisTracker-IR, is designed using a camera and an IMU. A 3D reconstruction process and incremental optimization are performed so that the coordinates of all LEDs are unified under the same world frame. The relative pose between the camera and IMU is calibrated using the method presented in (1). Then data from the IMU is fused with those from the vision-based estimation using the Extended Kalman Filter (EKF) to successfully compute the position and orientation of the camera in real-time. (3) A multiple IMUs based upper limb motion tracking system, LimbIMUTracker, is designed. An EKF-based IMU-segment relative orientation calibration is proposed. The position and orientation of the upper limb are computed in real-time via the calibration result, the output data of IMUs as well as the manually measured segment lengths. (4) ARTracker, a system that can simultaneously track both head and upper lim...
馆藏号XWLW1609
其他标识符200818014628070
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6360
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
杨浩. 面向增强现实的人体运动跟踪定位系统技术[D]. 中国科学院自动化研究所. 中国科学院研究生院,2011.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20081801462807(2662KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[杨浩]的文章
百度学术
百度学术中相似的文章
[杨浩]的文章
必应学术
必应学术中相似的文章
[杨浩]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。