基于运动图的运动合成方法研究
发布时间：2018-08-29 18:33
【摘要】： 虚拟人运动合成技术一直是虚拟现实领域研究的难点和热点之一,也是数字文化产业的核心技术之一,该技术在影视动漫、三维游戏、安全预演等诸多领域具有广阔的应用前景。在这些应用中不仅要求逼真的运动效果,而且要求能够灵活的控制虚拟人运动。本文针对这些应用需求,深入研究了基于运动图的运动合成方法,从运动图的数据准备、运动图的构建以及运动图的应用三方面展开,取得的主要研究成果如下: 一、在运动图的数据准备阶段,提出了一种基于特征的人体运动捕获数据自动分割方法。 对运动捕获数据进行合理的分割是构建运动图的前提和基础。对于分割得到的运动片断,一方面,运动片断的长度越短,对运动合成的控制越灵活;另一方面,为了方便用户控制,运动片断中应当包含明确的语义信息。我们利用运动中人体足部的相对关系提取运动特征,在此基础上,采用运动捕获数据与运动模板进行匹配的方式,实现了运动捕获数据的自动分割,避免了繁琐的手工操作,提高了运动捕获数据的分割效率和精度;同时,该方法具有很强的鲁棒性,运动分割的正确率比目前广泛使用的分割方法[KS05]提高了34.12%。 二、在运动图的构建阶段,提出了一种基于参数化运动合成的运动图方法。 运动图方法是一种有效地组织运动数据的方法,运动图的本质是一种由结点和边组成的有向图。目前运动图方法中存储的仅仅是少数运动捕获数据,这种方法使得运动图的表示能力有限,用户的控制精度不高。为此,我们提出了基于参数化运动合成的运动图方法,该方法的特点在于:其“结点”对应的是一个连续控制参数空间,给定一组控制参数,如运动速度、运动转角等,即可精确地合成出对应的运动片断,从而有效地提高了运动图的表示能力和用户的控制精度;其“边”表示了不同参数空间之间的转换关系,通过运动融合实现了运动片断的平滑过渡,同时,提出了一种根据约束计算根关节位置的方法,有效的避免了传统融合方法中脚步滑动和根关节朝向抖动的产生,提高了运动的逼真性。 三、面向运动图的具体应用,提出了一种基于路径的参数提取方法。 在上述构建的运动图中,控制参数是运动速度、运动转角等具有明确物理含义的参数值。然而,用户往往需要用直观的方式控制虚拟人运动,只给定运动轨迹,即要求能实时合成出一段满足轨迹要求的连续的运动序列。为了提供给用户一种灵活的交互方式,需要从直观的路径轨迹中提取出运动速度、运动转角等具体的运动参数。本文提出的基于路径的参数提取方法,既满足了运动合成中实时性的要求,又将路径轨迹和运动数据的属性紧密结合起来,使得合成的运动符合路径轨迹的要求。试验结果表明,该方法能够按照用户的轨迹要求实时地合成运动序列。
[Abstract]:Virtual human motion synthesis technology has always been one of the difficulties and hot spots in the field of virtual reality. It is also one of the core technologies of digital culture industry. Security preview and many other fields have broad application prospects. These applications require not only realistic motion effects, but also flexible control of virtual human motion. In this paper, the motion synthesis method based on motion graph is studied in order to meet the needs of these applications, including the preparation of motion graph data, the construction of motion graph and the application of motion graph. The main research results are as follows: firstly, a feature-based automatic segmentation method for human motion capture data is proposed in the data preparation stage of motion map. The reasonable segmentation of motion capture data is the premise and foundation of constructing motion graph. For the segmented motion segment, on the one hand, the shorter the length of the motion segment, the more flexible the control of motion synthesis; on the other hand, in order to facilitate user control, the motion segment should contain explicit semantic information. We use the relative relation of human feet in motion to extract motion features. On this basis, we use the matching method of motion capture data and motion template to realize the automatic segmentation of motion capture data and avoid the tedious manual operation. The efficiency and accuracy of motion acquisition data segmentation are improved, and the method has strong robustness, and the accuracy of motion segmentation is 34.1212% higher than that of the widely used segmentation method [KS05]. Secondly, a motion graph method based on parameterized motion synthesis is proposed in the construction stage of motion graph. Motion graph method is an effective method to organize motion data. The essence of motion graph is a directed graph composed of nodes and edges. At present, only a few motion capture data are stored in the motion graph method, which makes the representation of the motion graph limited and the user's control accuracy is not high. For this reason, we propose a motion graph method based on parameterized motion synthesis. The characteristic of this method is that its "node" corresponds to a continuous control parameter space, and a set of control parameters are given, such as motion velocity, motion rotation angle, etc. The corresponding motion segment can be synthesized accurately, which can effectively improve the expression ability of motion graph and the control accuracy of user, and its "edge" represents the conversion relationship between different parameter spaces. The smooth transition of motion segments is realized by motion fusion. At the same time, a method to calculate the position of root joints according to constraints is proposed, which effectively avoids the occurrence of step sliding and root joint orientation jitter in traditional fusion methods. The motion is more realistic. Thirdly, a path-based parameter extraction method is proposed for the application of motion graph. In the above motion diagram, the control parameters are the values of the motion velocity, the angle of motion and so on, which have definite physical meaning. However, users often need to control the motion of virtual human in an intuitive way. Only a given trajectory is given, that is, a continuous sequence of motion can be synthesized in real time to meet the requirements of trajectory. In order to provide users with a flexible way of interaction, it is necessary to extract the specific motion parameters, such as velocity and angle of motion, from the intuitionistic path. The path-based parameter extraction method proposed in this paper not only meets the real-time requirements of motion synthesis, but also closely combines the path and the attributes of motion data, making the synthesized motion meet the requirements of path trajectory. The experimental results show that the proposed method can synthesize motion sequences in real time according to the user's trajectory requirements.
【学位授予单位】：首都师范大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP391.41
本文编号：2212057
[Abstract]:Virtual human motion synthesis technology has always been one of the difficulties and hot spots in the field of virtual reality. It is also one of the core technologies of digital culture industry. Security preview and many other fields have broad application prospects. These applications require not only realistic motion effects, but also flexible control of virtual human motion. In this paper, the motion synthesis method based on motion graph is studied in order to meet the needs of these applications, including the preparation of motion graph data, the construction of motion graph and the application of motion graph. The main research results are as follows: firstly, a feature-based automatic segmentation method for human motion capture data is proposed in the data preparation stage of motion map. The reasonable segmentation of motion capture data is the premise and foundation of constructing motion graph. For the segmented motion segment, on the one hand, the shorter the length of the motion segment, the more flexible the control of motion synthesis; on the other hand, in order to facilitate user control, the motion segment should contain explicit semantic information. We use the relative relation of human feet in motion to extract motion features. On this basis, we use the matching method of motion capture data and motion template to realize the automatic segmentation of motion capture data and avoid the tedious manual operation. The efficiency and accuracy of motion acquisition data segmentation are improved, and the method has strong robustness, and the accuracy of motion segmentation is 34.1212% higher than that of the widely used segmentation method [KS05]. Secondly, a motion graph method based on parameterized motion synthesis is proposed in the construction stage of motion graph. Motion graph method is an effective method to organize motion data. The essence of motion graph is a directed graph composed of nodes and edges. At present, only a few motion capture data are stored in the motion graph method, which makes the representation of the motion graph limited and the user's control accuracy is not high. For this reason, we propose a motion graph method based on parameterized motion synthesis. The characteristic of this method is that its "node" corresponds to a continuous control parameter space, and a set of control parameters are given, such as motion velocity, motion rotation angle, etc. The corresponding motion segment can be synthesized accurately, which can effectively improve the expression ability of motion graph and the control accuracy of user, and its "edge" represents the conversion relationship between different parameter spaces. The smooth transition of motion segments is realized by motion fusion. At the same time, a method to calculate the position of root joints according to constraints is proposed, which effectively avoids the occurrence of step sliding and root joint orientation jitter in traditional fusion methods. The motion is more realistic. Thirdly, a path-based parameter extraction method is proposed for the application of motion graph. In the above motion diagram, the control parameters are the values of the motion velocity, the angle of motion and so on, which have definite physical meaning. However, users often need to control the motion of virtual human in an intuitive way. Only a given trajectory is given, that is, a continuous sequence of motion can be synthesized in real time to meet the requirements of trajectory. In order to provide users with a flexible way of interaction, it is necessary to extract the specific motion parameters, such as velocity and angle of motion, from the intuitionistic path. The path-based parameter extraction method proposed in this paper not only meets the real-time requirements of motion synthesis, but also closely combines the path and the attributes of motion data, making the synthesized motion meet the requirements of path trajectory. The experimental results show that the proposed method can synthesize motion sequences in real time according to the user's trajectory requirements.
【学位授予单位】：首都师范大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP391.41
【引证文献】
相关硕士学位论文 前2条
1 郭彬;基于运动捕获的角色运动合成研究与实现[D];电子科技大学;2010年
2 张兰兰;复杂虚拟场景的语义环境模型构建和人体动画生成[D];东北石油大学;2012年
，本文编号：2212057
本文链接：https://www.wllwen.com/wenyilunwen/dongmansheji/2212057.html
