ABSTRACT
Details of the measurement method should be referred to the original paper (Park and Hodgkin's, 2006). In this section, a summary of the method related to deformation modeling is mentioned. Static body shape of the reference pose is measured by a 3D scanner (FIGURE lea)). A total number of data points of the body surface is around one million. Then small mocap (motion capture) markers are located on the body surface like a mesh (FIGURE I (b)). The number of mocap markers is around 400. At that time, it is not required to place the mocap markers on anatomical landmarks. The reference pose is measured by a motion capture system. Then, the mesh connection information of mocap markers is defined manually. Moreover the near-rigid segment is defined as the mocap marker group according to the body segment (FIGURE I (c)). In the next process, dense data points measured by the 3D scanner are represented by sparse mocap markers. The near-rigid segments of mocap markers are moved to fit into the dense shape data automatically. Through this process, small difference of two reference poses between 3D scanning and motion capturing are corrected. After pose correction, dense data points are represented by the neighboring local coordinate system of mocap markers. One dense data point is represented by over three neighboring local coordinate systems of mocap markers. In motion sequence, the mocap makers are moved from the reference pose, then the dense point coordinate is moved according to the local coordinate systems of mocap markers based on initial representation of the reference pose. It is the base estimation of the deformation. Actually, mocap markers on the skin do not move rigidly. Thus, the remaining error caused by the skin deformation is modeled by a quadratic transformation matrix and the remammg error calculated from the model was added to the base estimation (FIGURE led)).
