基于视频素材的虚拟场景生成系统

发布时间：2018-10-12 08:18

【摘要】：虚拟场景正越来越多地被用于游戏娱乐、军事训练、广告与电影制作、旅游等各个领域。目前虚拟场景构建主要有两种方法,一种是通过三维建模技术来渲染生成,但是由于技术的限制,该方法暂时存在构建代价高、所需时间长,以及生成的虚拟场景不够真实等缺点；另一种是基于视频的虚拟场景构建,该方法具有素材丰富、细节真实等优点,但同时也存在处理困难等缺点。为了方便地利用已有视频素材来生成虚拟场景,我们需要一个高效的基于视频素材的虚拟场景生成系统。首先,本文分析了基于视频素材的虚拟场景构建系统所需的各项基本功能,并给出了各功能模块的设计。其中,为了高效地管理大量视频数据所需的视频素材管理系统,我们采用了一个层次化的视频数据关系模型,基于该模型以及视频切分、视频自动分类、视频对象提取技术,构建了我们的视频数据库,它能有效地管理视频素材,并支持视频的快速检索与浏览。本文还分析了系统各功能模块之间的接口及系统整合设计方案。其次,为了高效地生成虚拟场景,本文提出了视频融合过程中的前景对象编辑方法,包括调整对象的大小、旋转角度以及运动路径等参数,其中我们重点研究了怎样绘制对象的运动路径,从而能使得对象的运动与背景场景相融洽。接着,为了使得生成的虚拟场景更具有真实感,我们研究了对象视频存在抖动现象时会对视频融合产生的影响,并提出一个基于对象alpha视频轮廓曲线特征匹配的视频去抖动方法来消除对象视频的抖动,能得到比较好的去抖动效果。最后,本文给出了基于视频素材的虚拟场景生成系统实现,展示了其核心模块。该系统能够有效地管理视频数据,并能够方便地进行视频对象编辑与融合,生成虚拟场景视频。
[Abstract]:Virtual scenes are increasingly being used in games and entertainment, military training, advertising and film production, travel and other fields. At present, there are two main methods for virtual scene construction, one is rendering generation by 3D modeling technology, but because of the limitation of technology, this method temporarily has high construction cost and takes a long time. The other is the construction of virtual scene based on video, which has the advantages of rich material, real details and so on, but it also has some disadvantages such as difficulty in processing and so on. In order to make use of the existing video material to generate virtual scene conveniently, we need an efficient virtual scene generation system based on video material. Firstly, the basic functions of the virtual scene building system based on video material are analyzed, and the design of each function module is given. In order to efficiently manage the large amount of video data, we adopt a hierarchical video data relationship model, based on this model and video segmentation, video automatic classification, video object extraction technology. Our video database is constructed, which can effectively manage video material and support fast video retrieval and browsing. This paper also analyzes the interface between the various functional modules and the system integration design. Secondly, in order to generate virtual scene efficiently, this paper proposes a method for editing foreground objects in the process of video fusion, including adjusting object size, rotation angle and motion path, etc. We focus on how to draw the object's motion path so that the object's motion and background scene can be harmonious. Then, in order to make the generated virtual scene more realistic, we study the effect of the object video jitter on video fusion. A video dejitter method based on object alpha video contour curve feature matching is proposed to eliminate the jitter of object video. Finally, this paper presents the implementation of virtual scene generation system based on video material, and shows its core module. The system can manage video data effectively, edit and fuse video objects conveniently, and generate virtual scene video.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TP391.41

【参考文献】