reconstructing articulated shape of human hand and its skeletal structure from scanning

提出了一种通过观察人手在不同姿态下的三维形状来重建可摆姿态人手模型的方法。我们假设人手由18个刚性部件组成，并试图将残差建模为部件的非刚性变形。我们的技术从观察中恢复每个组件的3D形状。至于姿势操纵，我们的技术估计一个骨架，根据数据手套的传感器数据对齐组件。因此，我们可以重建一个真实的人手模型，并通过数据手套操纵其姿态。

We proposed a method of reconstructing posable human hand model by observing 3D shape of an individual hand in various postures. We assumed that a human hand consists of 18 rigid components and tried to model the residuals as non-rigid deformation of the components. Our technology recovers the 3D shape of each component from the observations. As for pose manipulation, our technology estimates a skeleton which aligns the components according to the sensor data of a data-glove. As a result, we could reconstruct a realistic human hand model and manipulate its posture via data-glove.