Research On Animation Synthesis Of Key Frames Based On 3D Motion Capture Data

With the development of science and technology,the application of animation synthesis technology is more and more extensive.People have a new functional requirement for the animation synthesis technology,which is to reuse the existing action to produce some different effects.That is,the animation technology is used to bring users a strong visual impact,as well as a shocking comedy effect.Nevertheless,the existing animation synthesis technology cannot meet this demand.To solve this problem,the research on the topic is studied in this thesis,and hopes to create a unique animation effect,with the purpose to create a unique animation effect through the existing action.The background and significance of the research is discussed in this thesis,as well as the existing problems of the research.Then a synthesis algorithm for 3D motion animation based on key frame is proposed.After focus on the relevant work of this thesis and describing the technology of animation synthesis,the technical features of motion synthesis and motion retargeting in the field of 3D animation synthesis are expounded.And the key technologies required for motion synthesis and motion retargeting are introduced,including the main segmentation of motion capture technology and key frame extraction technology.On the basis of theoretical research,two kinds of key frame extraction algorithms are put forward in this thesis.Finally,based on the algorithms proposed in this thesis,using the idea of interpolation,an animation synthesis system based on 3D motion capture data is designed and implemented.The main contents of this thesis are as follows:(1)A key frame extraction method of 3D motion capture data based on cosine distance hierarchical clustering is designed and implemented.This algorithm makes full use of the rotation of the joint points as the feature values of the motion capture data,then the high-dimensional data is mapped to low dimensional data by means of dimension reduction,after which the noise in the captured data is removed.Then the cosine distance represents the similarity degree and hierarchical clustering is used to split the data.Finally we put the segment points and the frames which has minimum error with means join the key frames sequence.The experiment shows that the proposed algorithm compared with the method based on rate and the method based on the curvature has high accuracy and recall rate.(2)The key frame extraction method of 3D motion capture data based on optimal segmentation is proposed.It is considered that the key frame extraction algorithm based on cosine distance hierarchical clustering is only based on the experience heuristic.The solution of the key frame is further expanded by us.Firstly,the model mechanism is used to build the model,and the model is used to model the motion capture data segmentation problem into an ordered sample clustering problem.Secondly,the optimal segmentation algorithm is used to segment the motion capture data.Finally,the median is extracted from each segment as the key frame.Finally,the segmentation points and the frame with the smallest mean error in each frame are extracted as the key frame.The experimental results show that,compared with DWT and PAA algorithm,the proposed method has better visual results,and the key frame has an absolute advantage in the expression of the original motion capture data,which can generalize and summarize the original motion capture data.(3)An animation synthesis system based on 3D motion capture data is designed and implemented.Based on the above two algorithms,combined with the knowledge of Bessel curve,using the Object-Oriented in the software design process,the animation synthesis system based on 3D motion capture data key frame is designed and implemented.The system not only has friendly interface,simple operation,complete functions,but also has high reusability of animation fragments,and can achieve a unique animation effect.And the availability of the algorithms are further verified.