Ontology-based Motion Graph Database And Application

The full life cycle automation of computer animation generation is a technology which converts limited natural language text into animation automatically. But in the technology, there is a bottleneck when generating motion and postures of the avatars: either manual editing with graphics software, or programming in assembly language is very laborious. One way to solve this problem is to use the motion database in animation software, for example, the popular 3D animation software named Maya. But the motions in the database tend to follow fixed formats, and the expressing ability is limited, so that they cannot be dynamically generated according to the needs of a story. Motion capture technology is the second way for the animators, but while the technology can generate various kinds of motions in a certain range, it is unable to meet the needs of automatically generating animation according to the circumstances of the story, because the motions are still generated beforehand. Motion graph technology which rises in recent years provides the possibility of using existing data to generate new motions, but it is not used in automatic animation generation research yet.In solving the above problems, this paper follow these guidelines: in order to more scientifically the connection between the storage avatar motion, we choose ontology as motion data organization carrier; in order to dynamically generated motion in accordance with the plot, we construct a motion graph database with scientific storage structure and rich content to support motion query, synthesis and interpolation; in order to utilize the limited data generated as much as possible of all kinds of motion, motion database stored in the motion should be modular for reusing the data in the greatest degree when synthesizing motion.When solving the problems above, this paper follows the following guidelines: constructing a motion graph database with scientific storage structure, rich content, and query function of the motion; In order to dynamically generate motion in accordance with the plot, the database needs to support the query of motion; Motion stored in the database should be modular for reusing the data in the greatest degree when synthesizing motion.Therefore, this paper did the works of the following several aspects:Firstly, identify and clipping of motion. Based on the original motion capture database provided by Carnegie Mellon University Graphics Lab, basic motion can be identify and clipped by lie group machine learning principle on the needs of the motion generating module in the system.Secondly, construct a motion ontology database. This paper represents and builds motion ontology database by OWL, and uses a data storage method with high efficiency and logical structure, namely uses OBDA to connect ontology and relational database. This makes motion clips not only have a clear logical structure, and can effectively query.Thirdly, query and synthesize motion. Based on ontology Editing tools Protégé and ontology query language SPARQL, according to the need of motion synthesis for different scene and avatar constraints, query the motion in database, and synthesize motion clips.Fourthly, combine path-and-posture planning in 3D environment. To test the ability of the motion database to solve practical problems, this paper design a algorithm of combined path and posture planning, and implement a simple system that avatar automatically choose the best path and postures through different 3D environment. After testing on different scenes, the system can generate avatar motions that conform to the constraints.