基于三维重建的民族舞蹈展示系统设计与实现

发布时间：2018-03-20 16:42

本文选题：三维重建　切入点：民族舞蹈　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着传统民族文化传播形式的不断发展,传统的以文字、图片为主要表现形式的传播方式已不能满足人们的需求,而虚拟数字化代表着未来发展的方向。针对此问题,本文主要研究民族舞蹈的三维可视化,从而使民族舞蹈实现数字化的目的。本文首先提出了一种基于面片的多视立体重建解决方案,利用此方案可以精确的重建三维角色模型。然后利用运动捕捉系统采集民族舞蹈运动数据,驱动三维角色模型产生舞蹈展示动画。本文的研究工作主要包括以下几个部分:(1)数据获取与姿态估计,数据获取包括二维原始数据的获取和民族舞蹈的BVH数据的获取。前者根据国内外学者提供的部分标准,通过实验不断修正拍摄方式,最终提出标准化的拍摄方案;后者通过运动捕捉系统对民族舞蹈家的舞蹈动作进行采集,并通过一系列处理转换成BVH数据。本文采用基于SIFT的图像匹配方法,在传统的SIFT特征向量匹配基础上加入Ransac算法消除噪音。使用流行的基于Levenberg-Marquardt算法的Bundler,得到成像摄像机运动参数、场景的结构信息。(2)点云融合和渲染,本文采用基于面片的多视立体的点云融合算法实现人物点云模型的构建。用三维Delaunay三角剖分算法替代传统的泊松重建算法生成三角网格模型。通过多视立体重建算法获得的三角网格模型比较粗糙,本文采用借助Zbrush的ZRemesher和Geometry-Divide功能实现模型的平滑和简化。并利用动态贴图的方法对三角网格模型进行纹理映射,从而使模型更加逼真。(3)民族舞蹈系统展示,利用运动捕捉获得的BVH数据驱动人物模型产生民族舞蹈动画。研究内容包括:人物骨骼的建立,BVH数据的导入。本文采用基于距离场的自动生成骨骼算法完成民族舞蹈系统的实现。最终基于三维重建的民族舞蹈在Maya中展示。实验结果表明,本文实现的民族舞蹈展示系统能够满足将民族舞蹈数字化的目的,并且操作简单,成本低廉,可视化效果好。
[Abstract]:With the continuous development of the traditional national culture communication, the traditional communication with text and picture as the main form of communication can no longer meet the needs of people, but virtual digitization represents the direction of future development. In this paper, the three-dimensional visualization of folk dance is mainly studied, so that the national dance can be digitized. Firstly, a multi-view stereo reconstruction solution based on facial slices is proposed. Using this scheme, 3D role model can be reconstructed accurately, and then the motion capture system is used to collect the data of folk dance movement. The research work of this paper mainly includes the following parts: 1) data acquisition and attitude estimation. Data acquisition includes the acquisition of two-dimensional raw data and the acquisition of BVH data of folk dance. The former, according to some standards provided by scholars at home and abroad, constantly modifies the shooting methods through experiments, and finally puts forward a standardized shooting scheme. The latter collects the dance movements of the folk dancers through the motion capture system and converts them into BVH data through a series of processing. This paper adopts the method of image matching based on SIFT. Based on the traditional SIFT feature vector matching, Ransac algorithm is added to eliminate noise. By using the popular Levenberg-Marquardt algorithm, the motion parameters of the imaging camera and the structure information of the scene are fused and rendered. In this paper, the point cloud fusion algorithm based on multi-view stereo is used to construct the point cloud model, and the three-dimension Delaunay triangulation algorithm is used to replace the traditional Poisson reconstruction algorithm to generate the triangular mesh model. The triangular mesh model obtained by the algorithm is rough. In this paper, the ZRemesher and Geometry-Divide functions of Zbrush are used to smooth and simplify the model, and the method of dynamic mapping is used to map the texture of the triangular mesh model, thus making the model more realistic. The BVH data from motion capture is used to drive the character model to produce the folk dance animation. The research content includes: the establishment of the character skeleton and the import of the BVH data. In this paper, the automatic bone generation algorithm based on the distance field is used to complete the ethnic dance animation. The realization of dance system. Finally, the national dance based on 3D reconstruction is displayed in Maya. The experimental results show that, The national dance display system realized in this paper can meet the purpose of digitizing the national dance, and the operation is simple, the cost is low, and the visualization effect is good.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;J722.22

【参考文献】