基于头盔显示器的视频头位跟踪方法研究

Video-based Head Tracking with HMD

头位跟踪在虚拟现实系统中起着至关重要的作用。提出了一种新的基于头盔显示器的视频头位跟踪方法。首先设计了立体标记物,置于头盔显示器正上方,然后使用CAMSHIFT算法跟踪视频中标记物的四个非共面特征点,获取其像素坐标,跟踪过程中使用OpenMP实现了多点的同步跟踪。最后利用非共面的POSIT算法估计立体标记物的旋转矩阵和平移向量,得到头部位姿。实验表明,该方法是有效可行的,精度高、实时性好、无束缚。

Head tracking plays an important role in the VR （Virtual Reality） system. A novel video-based real time head tracking method with HMD was proposed. Firstly, a special marker laid on the HMD was designed for the system. CAMSHIFT algorithm was then employed to get the pixel positions of the corresponding feature points in the single-view video sequence. To overcome the efficiency bottleneck, OpenMP was employed for multiple targets tracking. Finally, the rotation matrix and translation vector of the marker were estimated by using the non-coplanar POSIT algorithm, and the pose of the head was attained from the rotation matrix. Experiments show the proposed method is feasible and effective with high accuracy and rate, and has no constraint.