中文摘要：

为了更逼真地从视频图像序列中实现三维人体骨架动画形式的提取，以便进一步地对人体运动进行分析与研究，提出了一种基于多视角视频的运动重建的方法。该方法充分利用了标记点的信息，其核心步骤有标定摄像机，提取标记点，跟踪标记点和人体运动三维重建四个主要方面。其中，在跟踪标记点时，使用了基于多视觉的目标跟踪算法，该算法由结合了扩展卡尔曼滤波预测与标记点轨迹平滑性约束所构成的双目立体视觉跟踪与多目视觉数据融合两个方面。实验结果证明了所提方法的有效性与可行性。

英文摘要：

In order to realistically realize 3 D human skeleton animation from image sequences of video, for the purpose of researching and analyzing human movement further, a method of human reconstruction based on multi-view is presented. The method makes full use of the markers＇ information, of which the key technologies include camera calibration, markers extraction, markers tracking and human motion 3D reconstruction. In the procedure of tracking markers ,propose a novel multi-view tracking algorithm. This algorithm is made up of two steps, one is binocular stereo vision tracking which combines the Extended Kalman Filter with the smooth of markers＇ trajectory, and another is multi-view data fusion. The effectiveness and feasibility of this method is verified by experiments.