The Design And Implementation Of Skinned Mesh Character Animation Based On Motion Capture Data

In recent years, with the quick development of the software and hardware of computer technology, various 3D virtual character animations and games are gradually playing more and more important roles in human daily life and entertainment. Computer based virtual character animation technology is now being considered as the core technology of computer animation technology, virtual reality and augmented reality. The simulation of the real movements and behaviors of human in the real-world through computer based 3D virtual characters, that is, the dynamic simulation technology of 3D virtual characters, is now becoming the hot top in both academic and industrial fields. There are three key issues for how to simulate human motions realistically, namely a) modelling of 3D virtual characters, b) controlling of motions and behaviors, c) binding of bone and skin mesh. In the circumstances, this thesis focuses on the research of the modeling technology of 3D virtual characters, and employs motion capture data ASF/AMC files for driving the established virtual character models, obtaining the animation and simulation effects.1) Both the ASF and AMC files of the motion capture data are analyzed and explained in detail. Wherein, the ASF file defines the initial state of the bone model of a 3D virtual character. Based on ASF, this thesis builds up the virtual character bone model, which consists of 31 joints and the corresponding 30 bones. The topological structure of the joints and bones is further established. On the other hand, the AMC file contains the frame sequential of the virtual character. Besides, this thesis also builds up the skin model of the virtual character, which is made up by a polygon mesh model containing numerous skin points with special topologic relations. The skin model is further texture mapped in order to make it much more verisimilar.2) This thesis proposes a weight computational method for computing the weights of the virtual character skin points. After matching the established bone model and skin model, this thesis applying the idea of bone bounding box to divide the affect regions of each bone on the skin mesh. Therefore, the weights of the skin points can be calculated.3) The skinning method of the polygon mesh based skin model is researched in detail, including two algorithms, namely linear blending skinning(LBS) and dual quaternions blending skinning(DQBS), to realize skinned mesh character animation. The LBS algorithms is firstly applied. When driven by the motion capture data ASF/AMC, some skin distortions will happen such as skin “collapse” and “knotting”, etc. The reason of these distortions is that the resulting matrix computed by the linear blending of the weighted rigid transformation matrix is not rigid any more, which cannot keep the volume of the skin model. Therefore, this thesis applies the DQBS algorithm to improve the LBS algorithm, and the comparison of the two algorithms is then conducted too.4) Finally, a simulation program is developed through C++ in.NET environment. The skin deformation experiments are conducted with the motion capture data ASF/AMC files by the program. The results shows that the both the established bone model and the skin model of the 3D virtual characters are practical, and can obtain better animation and simulation by applying DQBS algorithm.