基于Kinect网络的实时动作捕捉与比对系统的设计与实现

发布时间：2018-07-03 00:23

本文选题：动作捕捉 + 对比分析　；参考：《山东大学》2017年硕士论文

【摘要】：动作捕捉在游戏、动漫、教育培训等领域有着广泛的应用。目前,随着虚拟现实(Virtual Reality,VR)技术的蓬勃发展,实时动作捕捉也成为众多VR系统的必备组成部分,支持用户与VR内容进行自然交互。其中,微软公司2010年推出的动作捕捉设备Kinect,因其成本比较低、便携、运行可靠等优点而得到广泛应用。但是目前的Kinect捕获范围有一定限制,而且只能较好对用户的正面动作进行捕捉。当用户侧身朝向Kinect时,Kinect无法正常采集用户骨骼关节点数据,从而限制了其应用范围。本文研究背景是虚拟太极拳学习环境的建设,需要全方位捕捉与再现太极拳学习者的动作,并与标准动作进行比对分析。为此,本文设计和开发了一个基于多Kinect的360度全方位动作捕捉系统。本文通过研究分析已有的动作捕捉系统,尤其是基于光学的动作捕捉系统,并结合Kinect的技术特性,在利用传统光学摄像机校正算法的基础之上,融合多Kinect之间骨架数据接力算法和Kinect数据的滤波算法,研发了一个基于Kinect网络的动作捕捉和对比分析系统。其中,动作对比分析用于两组动作数据之间的对比,给出不符合标准的关节点改进意见,并且给出整个动作的一个量化的标准度分数。系统主体部分采用Unity3D进行开发渲染;Kinect数据服务器端是WPF工程进行开发的;而校正计算部分使用MatLab完成;不同技术模块间则通过网络通信的方式联系在一起构成整个系统。针对不同的任务需求使用不同的技术进行开发,提高了开发速度和可维护性。
[Abstract]:Action capture in the game, animation, education and training has a wide range of applications. At present, with the rapid development of virtual reality (VR) technology, real-time motion capture has become an essential part of many VR systems, supporting users to interact with VR content naturally. Among them, Kinect, a motion capture device introduced by Microsoft in 2010, has been widely used because of its advantages of low cost, portable and reliable operation. However, the current Kinect capture range is limited, and can only capture the positive actions of users. When the user is sideways facing Kinect, Kinect can not collect user's skeleton node data normally, which limits its application range. The research background of this paper is the construction of virtual Taijiquan learning environment, which needs to capture and reproduce the actions of Taijiquan learners in all directions, and compare and analyze with the standard movements. Therefore, this paper designs and develops a 360-degree motion capture system based on multi-Kinect. By studying and analyzing the existing motion capture system, especially the motion capture system based on optics, and combining the technical characteristics of Kinect, this paper makes use of the traditional optical camera correction algorithm. A motion capture and contrast analysis system based on Kinect network is developed by combining the relay algorithm of skeleton data between multi-Kinect and the filtering algorithm of Kinect data. The action contrast analysis is used to compare the two groups of action data, and the improvement advice of the non-standard gate node is given, and a quantized standard score of the whole action is given. The main part of the system uses Unity3D to develop and render Kinect data server, which is developed by WPF project, while the correction calculation part is completed by MATLAB, and the different technical modules are connected together through the way of network communication to form the whole system. The development speed and maintainability are improved by using different technologies to meet different task requirements.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.9

【参考文献】