DefinePK

Abstract

In 3D character animation, the creation of the skeleton is still long-winded task and relies on various manual modifications. In this paper, a novel approach is proposed to generate a kinematic skeleton for 3D human geometric model. More specifically, our concentration to create the plausible skeleton, which locates the precise position of the knee and elbow joints of the models that possessing straight limbs. Since inaccurate identification of the joints in skeleton extraction effected the quality of the character animation. Firstly, in the proposed approach, approximate locations of joints in the static model are captured. The Laplacian-based mesh contraction method with global positional constraints has been applied to obtain the skeletal shape of the model. Secondly, the correspondence relations between the joints and contracted mesh are established through K-nearest neighbors. Finally, some experiments have been performed to generate plausible kinematic skeletons on different 3D models. In addition, the correctness of the generated skeleton has been assessed by animating the skeleton. The animation of the obtained skeleton carried out through joint mapping by using 3D motion. The resulting skeletons could find its applications in 3D shape corresponding, shape interpolation mesh skinning, character animation and mesh editing.